Module 1: Introduction to AI and Prompt Engineering
Evolution of AI in media
Basics of NLP and LLMs
What is a prompt?
The role of prompt engineering in digital storytelling
Introduction to AI and Prompt Engineering: A New Era in Media
In the last few years, media has witnessed a revolutionary transformation, with Artificial Intelligence (AI) no longer playing a behind-the-scenes role. From newsroom automation to synthetic news anchors, from AI-generated headlines to deepfake documentaries, the disruption is real and rapidly evolving. At the heart of this transformation lies a powerful, but often overlooked, tool: the prompt.
A "prompt" is no longer just a question or command—it’s the key that unlocks the immense potential of generative AI systems. And "prompt engineering" is the art and science of crafting these keys with precision, clarity, and creativity. As journalism, filmmaking, podcasting, and social media adapt to this new world, understanding prompt engineering is becoming essential for every media professional.
This chapter explores the evolution of AI in media, the basics of Natural Language Processing (NLP) and Large Language Models (LLMs), what prompts actually are, and how prompt engineering is shaping the future of digital storytelling.
The Evolution of AI in Media
AI’s integration into the media ecosystem has not happened overnight. Its journey can be traced back decades, starting from rule-based systems and robotic journalism to the latest wave of generative AI.
Stage 1: Rule-based Systems and Early Automation (1980s–2000s)
In the early days, AI in media was mostly about automation. News agencies like Reuters and Bloomberg experimented with templates that could generate earnings reports or sports summaries by feeding in structured data. These were rule-based AI systems, where human programmers wrote specific logic and instructions.
These systems worked well in domains where the input data was predictable and structured. However, they lacked creativity, adaptability, or linguistic nuance.
Stage 2: Machine Learning and News Recommendation (2000s–2015)
With the growth of machine learning, media companies began using AI to understand audience preferences. Netflix’s recommendation algorithm, Facebook’s news feed, and Google News curation are examples. These were systems trained on user behavior, using pattern recognition to tailor content delivery.
During this phase, AI was mainly used behind the scenes—for content personalization, programmatic advertising, and basic summarization. It wasn’t creating the content yet.
Stage 3: NLP and Content Assistance (2015–2022)
The emergence of Natural Language Processing (NLP) models like BERT (by Google) and GPT-2 (by OpenAI) brought AI closer to language understanding. Tools like Grammarly, Wordtune, and AI transcription software like Otter.ai began helping journalists and writers with grammar, tone, and efficiency.
Media houses started using AI to assist in:
Transcribing interviews
Generating metadata
Translating stories into local languages
Flagging misinformation or plagiarism
Stage 4: Generative AI and the Rise of LLMs (2022 onwards)
The release of ChatGPT in late 2022, based on OpenAI’s GPT-3.5 and later GPT-4, triggered the generative AI revolution. Suddenly, AI wasn’t just correcting grammar or recommending headlines—it was writing articles, designing layouts, creating visuals, composing music, and even mimicking human voices.
In India, newsrooms like The Indian Express, The Print, and Mathrubhumi began experimenting with LLMs for:
Drafting articles
Summarising news reports
Creating explainers and FAQs
Translating between English and regional languages
Designing AI-generated visuals for news graphics
It became clear: AI was no longer an assistant—it was now a co-creator. And prompts were the language we needed to speak to it.
Basics of NLP and LLMs
To understand how prompts work, we need a basic grasp of Natural Language Processing (NLP) and Large Language Models (LLMs).
What is NLP?
NLP is a subfield of Artificial Intelligence that deals with the interaction between computers and human language. It involves teaching machines to understand, interpret, generate, and respond to human language in a meaningful way.
Some common NLP tasks include:
Speech recognition (e.g., Siri, Google Assistant)
Sentiment analysis (e.g., understanding tweets)
Machine translation (e.g., Google Translate)
Summarisation (e.g., news abstracts)
Named Entity Recognition (identifying people, places, etc.)
Question Answering (e.g., ChatGPT)
NLP combines linguistics, statistics, and machine learning to bridge the gap between humans and machines.
What are Large Language Models (LLMs)?
LLMs are deep learning models trained on massive amounts of text data. They are designed to predict the next word in a sentence. Over time, this basic ability has been scaled to such an extent that these models can now generate full articles, write code, create poetry, and answer complex questions.
Popular LLMs include:
GPT (OpenAI) – Used in ChatGPT
Claude (Anthropic)
Gemini (Google)
LLaMA (Meta)
Mistral, Cohere, Command-R, etc.
These models are trained on everything from Wikipedia to books, news articles, social media posts, and code repositories. The more diverse and large the training data, the better the model’s ability to generate coherent and context-rich responses.
LLMs do not "think" like humans. They work based on probability—calculating the most likely sequence of words based on a given input (prompt). That’s why how we ask or instruct them matters immensely.
What is a Prompt?
In simple terms, a prompt is the input you give to an AI model to get a desired output. It’s like asking a question, giving an instruction, or setting a task.
For example:
Prompt: “Write a 200-word summary of this news article.”
Prompt: “Translate the following paragraph into Malayalam.”
Prompt: “Give me five Instagram caption ideas for World Photography Day.”
Just like the way a journalist frames a question in an interview to get the best quote, crafting a prompt is about being clear, focused, and intentional.
Prompts can be:
Instructional: "Explain the concept of climate change."
Conversational: "Hi! I’m preparing a report. Can you help summarize this?"
Role-based: "You are a senior editor at The Hindu. Rewrite this intro with a formal tone."
Few-shot: Giving the AI a few examples before asking for a new one.
The better your prompt, the better the result. Vague or poorly structured prompts often yield unhelpful or incorrect outputs.
The Role of Prompt Engineering in Digital Storytelling
Prompt Engineering is the practice of designing, refining, and optimizing prompts to guide AI systems in generating accurate, creative, and context-aware responses.
In the media world, prompt engineering is the new writing skill. It’s how you direct AI to:
Create a compelling headline
Generate social media hooks
Visualize a news story using image generators
Summarise long reports into digestible tweets
Translate content while retaining emotion and tone
Assist in scriptwriting or podcast episode planning
Here’s how prompt engineering is already reshaping digital storytelling:
1. Speed and Productivity
With well-designed prompts, journalists can generate story drafts in minutes. Editors can quickly test multiple headlines. Translators can work faster with AI assistance. This frees up time for investigation, fact-checking, and analysis.
2. Creativity Enhancement
Rather than replacing creative professionals, prompt engineering acts as a creativity booster. A filmmaker can prompt DALL·E to visualise storyboards. A podcaster can brainstorm show titles with GPT. A writer can unblock writer’s block.
3. Personalization at Scale
AI tools prompted with audience data can help create hyper-personalized news digests, custom video summaries, or community-specific explainers. This is particularly useful in platforms like WhatsApp, Instagram, and Threads.
4. Language Diversity
With prompts, LLMs can be guided to translate, summarize, or rewrite content in regional languages. This democratizes access to news and information, especially in a country like India with 22 official languages and hundreds of dialects.
5. Cross-Format Storytelling
A single prompt can be used to generate multiple formats:
A 1000-word article
A 1-minute YouTube script
3 image prompts for a social media carousel
A 30-second podcast ad
This makes multimedia storytelling faster and more consistent.
Challenges and Responsibilities
While the power of prompt engineering is evident, it also comes with risks and ethical responsibilities.
Hallucinations: LLMs can generate believable but false content.
Bias: If prompted incorrectly, AI can reinforce stereotypes or political biases.
Plagiarism: AI-generated content may accidentally echo copyrighted materials.
Transparency: Media houses must disclose AI usage to maintain audience trust.
That’s why learning prompt engineering is not just about writing better commands—it’s about understanding the AI’s limitations, verifying outputs, and applying editorial judgment.
Why This Matters Now
AI is here. It’s not a passing trend. For media professionals, ignoring it is like ignoring the printing press in 1440 or the internet in 1996.
Prompt engineering is the bridge between human creativity and machine intelligence. It empowers journalists, content creators, and communicators to direct AI in ethical, accurate, and impactful ways.
This course will teach you how to think like a prompt engineer, experiment with various tools, and master this emerging skill that will define the future of media.
Because in the age of AI, the prompt is the new pen.
Module 2: Understanding Generative AI Tools for Media
Overview of tools: ChatGPT, Claude, Gemini, Copilot, DALL·E, Midjourney, RunwayML, Descript, etc.
Text, image, audio, and video generation
Tool selection based on media goals
Understanding Generative AI Tools for Media
A New Creative Partner for the Media Industry
Imagine a newsroom where headlines write themselves, a podcast team that creates voiceovers with a single click, and a social media editor who can generate images that mirror global trends within seconds. Welcome to the era of Generative AI—a set of technologies that can produce text, audio, visuals, and even videos based on written prompts.
For the media industry, Generative AI has transformed workflows, creative practices, and even audience engagement. Whether you’re a journalist drafting breaking news, a filmmaker visualizing a storyboard, or a social media manager looking for viral captions, there’s a tool to help you. But the challenge lies not in availability—it lies in choosing the right tool for the right purpose, and knowing how to interact with it effectively.
This module explores the major generative AI tools—ChatGPT, Claude, Gemini, Copilot, DALL·E, Midjourney, RunwayML, Descript, and more—across their modalities: text, image, audio, and video. We also learn how to select the right tool based on specific media goals.
What Is Generative AI?
Generative AI refers to algorithms that can create new content—such as text, images, music, or video—based on patterns learned from vast datasets. These models use machine learning to “generate” rather than just analyze or retrieve data.
There are four main types of generative outputs relevant to the media:
Text: News stories, captions, summaries, scripts.
Image: Illustrations, posters, thumbnails, memes.
Audio: Voiceovers, podcasts, background music.
Video: Edits, transitions, explainer videos, talking avatars.
Each of these content types is powered by different tools and models, and each has specific use cases in journalism, content creation, and digital storytelling.
Text Generation Tools: For Writers, Editors, and Researchers
ChatGPT (OpenAI)
Best for: Writing, editing, research, brainstorming.
ChatGPT is the most widely used text-based AI tool. It runs on OpenAI’s GPT (Generative Pre-trained Transformer) models—GPT-3.5 (free version) and GPT-4 (Pro version). ChatGPT can write articles, summarize content, generate headlines, translate languages, suggest interview questions, and assist in creative writing.
Media use cases:
Drafting articles from bullet points.
Explaining complex topics in simple language.
Translating between Indian languages.
Generating SEO-friendly headlines and intros.
Fact-checking and paraphrasing content.
Strengths:
Highly conversational.
Supports web browsing in Pro mode.
Integrated with image tools (DALL·E).
Claude (Anthropic)
Best for: Long document understanding and summarisation.
Claude is designed to be “helpful, harmless, and honest.” Its strength lies in processing large files (over 100,000 tokens) and maintaining context in long conversations.
Media use cases:
Summarising PDF reports and transcripts.
Drafting long-form features.
Brainstorming media campaign ideas with detailed back-and-forth.
Strengths:
Understands nuance.
More concise than ChatGPT in long summaries.
Better at retaining context in structured documents.
Gemini (Google)
Best for: Google-integrated workflows, real-time data.
Previously called Bard, Gemini is Google’s AI chatbot that leverages live data and integrates with Google Workspace tools.
Media use cases:
Quick news analysis from live data.
Drafting with references and sources.
Integrating AI into Google Docs, Sheets, or Gmail.
Strengths:
Real-time data integration.
Native search accuracy.
Seamless with Android and Google Drive.
Copilot (Microsoft/OpenAI)
Best for: Office users, enterprise environments.
Copilot is Microsoft’s integration of OpenAI tools within Word, PowerPoint, Excel, and Outlook. It can auto-generate slides, summarize emails, or help write scripts in PowerPoint.
Media use cases:
Creating newsroom reports.
Auto-generating social media campaign summaries.
Making decks for editorial pitches.
Strengths:
Works inside Office apps.
Enterprise-grade security.
Great for automation-heavy environments.
Image Generation Tools: For Designers and Visual Storytellers
DALL·E (OpenAI)
Best for: News illustrations, concept art, explainer visuals.
DALL·E can generate photorealistic or artistic images from a text prompt. Integrated within ChatGPT (Pro version), it also allows inpainting—editing images by describing changes.
Media use cases:
Creating thumbnails for articles or YouTube.
Generating images for events (e.g., World Press Freedom Day).
Designing editorial illustrations.
Strengths:
Easy to use with ChatGPT.
Editable with prompt-based refinements.
Fast and accessible.
Midjourney
Best for: Artistic and cinematic images.
Midjourney produces high-quality, surreal or artistic images using prompts sent via Discord. Its outputs often appear stylized and are favored in magazines, posters, and creative campaigns.
Media use cases:
Editorial illustrations with emotion.
Magazine covers.
Visual branding for campaigns.
Strengths:
Hyper-stylised outputs.
Community-generated prompt sharing.
Useful for visual metaphors.
Ideogram
Best for: Text-on-image generation.
Ideogram allows you to create visuals with text embedded in them—great for typographic posters, signs, or graphic posts.
Media use cases:
Making quote cards.
Social media slides.
Digital posters with call-to-actions.
Strengths:
Clear text output.
Poster-ready designs.
Fast iteration.
Audio Tools: For Podcasters, Voiceovers, and Audio Storytellers
Descript
Best for: Podcast editing, transcription, voice cloning.
Descript allows users to edit audio and video like a Word document. It automatically transcribes interviews, removes filler words, and enables voice replacement using AI.
Media use cases:
Creating podcasts and news audio summaries.
Editing interviews with auto-cuts.
Adding narration to explainers.
Strengths:
Simple UI.
Multitrack editing.
Overdub (voice cloning) feature.
ElevenLabs
Best for: Hyper-realistic voice synthesis.
ElevenLabs can generate lifelike voiceovers from text. Users can choose from stock voices or clone their own voice for narration or multilingual voiceovers.
Media use cases:
AI voiceovers for YouTube news explainers.
Dubbing regional language content.
Audio storytelling for documentaries.
Strengths:
Natural intonation.
Emotion and tone control.
Fast processing.
Play.ht
Best for: TTS (Text to Speech) for multiple accents.
Play.ht specialises in converting scripts into professional-grade voiceovers with a variety of accents and tones.
Media use cases:
Creating international news bulletins.
Narrating blogs or written articles.
Making IVR-like media bots.
Strengths:
High voice variety.
Simple web interface.
Supports SSML markup.
Video Tools: For Visual Editors and Content Producers
RunwayML
Best for: AI video editing, motion graphics, text-to-video.
Runway is a powerful video tool for creatives who want to automate video edits, generate B-roll, or experiment with AI animations.
Media use cases:
AI-powered video editing for news packages.
Converting scripts into AI visuals.
Removing backgrounds or objects in interviews.
Strengths:
Visual storytelling with minimal manual input.
Text-to-video capability.
Ideal for social media shorts.
Pika Labs
Best for: Experimental and animated video generation.
Pika Labs allows creators to generate video scenes based on text prompts. It’s still experimental but fast-growing in community use.
Media use cases:
Creating animation explainers.
Adding dynamic effects to shorts.
Storyboarding ideas for docu-films.
Strengths:
Fast outputs.
Stylised animations.
Good for creative workflows.
Tool Selection Based on Media Goals
Choosing the right AI tool depends on what you want to achieve in your content workflow. Here's a breakdown by goal:
✅ For Reporting and Writing
Use: ChatGPT, Claude, Gemini
Tasks: Drafting stories, summarizing, writing FAQs, checking facts
Why: These tools are optimized for coherent, contextual, and reference-rich outputs.
✅ For Headline Creation and SEO
Use: ChatGPT, Copilot
Tasks: Headlines, metadata, keywords
Why: LLMs can A/B test different headline styles and generate audience-targeted options.
✅ For Visual Communication
Use: DALL·E, Midjourney, Ideogram
Tasks: Illustrations, thumbnails, social media creatives
Why: Fast image generation without depending on stock photos.
✅ For Podcasts and Audio Storytelling
Use: Descript, ElevenLabs, Play.ht
Tasks: Interview editing, narration, translations
Why: Speeds up production, reduces dependency on voice artists.
✅ For Social Media Campaigns
Use: ChatGPT + DALL·E + Descript
Tasks: Captions, memes, reels
Why: Cross-format content in one flow with visual and audio integration.
✅ For Explainers and Video Editing
Use: RunwayML, Pika Labs
Tasks: B-roll creation, animation, AI-assisted editing
Why: Saves hours of video production and adds dynamic effects.
Key Considerations Before Choosing a Tool
Output Format – Do you need text, voice, video, or image?
Creative Control – Do you want to steer the design or just generate?
Collaboration – Is it for solo use or teamwork?
Language and Region – Does it support multilingual or Indian context?
Cost and Licensing – What is the pricing model and content rights?
Tools Don't Replace You—They Multiply You
Generative AI tools are not here to replace journalists, editors, or creators. They’re here to amplify your creativity, accelerate production, and enable multimedia storytelling.
The key is not just knowing what tool exists—but knowing how and when to use it. As you progress in this course, you’ll learn how to prompt these tools effectively, combine them for cross-media workflows, and adapt them responsibly to your editorial mission.
In the age of AI, the tools are ready. The creativity still comes from you.
Module 3: Types of Prompts and Media Use Cases
Informational vs Creative prompts
Instructional, conversational, zero-shot and few-shot prompts
Case studies from media (Reuters, NYT, Indian Express, The Hindu, BBC)
Why the Type of Prompt Matters
In the world of generative AI, especially for media professionals, prompts are not just commands. They are creative levers, editorial tools, and strategic assets. The type of prompt you use can entirely shape the tone, depth, format, and accuracy of the AI-generated content.
Think of prompt engineering as journalism in reverse. Instead of asking a source and extracting information, you’re instructing a machine on how to produce content—fact-based or imaginative, concise or detailed, professional or casual. Whether you’re drafting a tweet, writing a news explainer, or generating a visual, the form and function of your prompt is everything.
In this module, we explore:
The difference between informational and creative prompts.
Key prompting frameworks like instructional, conversational, zero-shot, and few-shot.
Real-world case studies from media giants like Reuters, The New York Times, Indian Express, The Hindu, and BBC that show how prompt types influence AI output quality and usefulness.
Let’s decode the mechanics and strategy of effective prompting in journalism and content creation.
1. Informational vs Creative Prompts
AI tools like ChatGPT, Claude, Gemini, or Copilot can produce both factual and imaginative outputs—but the intention behind your prompt must be clear. Broadly, prompts can be categorized into two families:
✅ Informational Prompts
These prompts seek facts, summaries, explanations, or clarifications. They rely on knowledge-based outputs, often rooted in real-world context.
Examples:
“Summarise today’s Lok Sabha debate in 200 words.”
“Explain the economic implications of the RBI interest rate hike.”
“What are the key points from the IPCC climate report 2025?”
These prompts are used for:
News summaries
Report writing
FAQs
Explain-it-like-I’m-five breakdowns
Output Qualities: Objective, structured, clear, concise
Used by: Reporters, copyeditors, research desks
Creative Prompts
These prompts instruct the AI to invent, reimagine, or simulate scenarios. They're useful for marketing, audience engagement, and brand storytelling.
Examples:
“Write a poem about press freedom as if spoken by a typewriter.”
“Create a tweet thread imagining how Mahatma Gandhi would view WhatsApp.”
“Design an Instagram caption for a photo of a flooded Kerala road.”
Used for:
Social media campaigns
Headlines and hooks
Fictionalised intros
Visual storytelling
Output Qualities: Imaginative, stylistic, engaging, subjective
Used by: Content creators, social media teams, visual editors
Key Difference:
Informational prompts retrieve and synthesize existing knowledge.
Creative prompts generate new ideas or perspectives using language and stylistic mimicry.
A well-rounded journalist or media professional should know when to switch between these two modes, depending on the audience, platform, and editorial intention.
2. Prompting Frameworks
Now let’s break down how prompts can be designed, beyond just the informational/creative divide. There are several frameworks widely used in the industry:
Instructional Prompts
These are direct commands or tasks given to the AI.
Structure:
"Write a news brief about X in Y words."
"Translate the following to Malayalam."
"List five key takeaways from this press release."
Use Case Example:
A newsroom uses the prompt:
“Write a 100-word summary of the G20 Summit highlights for Instagram caption format.”
Best For:
Summarization
Translation
Headline generation
Data transformation
Tip: Be specific with word count, tone, and format.
Conversational Prompts
These prompts simulate a dialogue, often casual or exploratory in tone. They allow iterative refinement and back-and-forth.
Structure:
“Hi! I’m working on a news explainer. Can you help simplify this?”
“I want to understand the background of the Hamas-Israel conflict. Start with 1948.”
“Can you act like a senior editor at The Hindu and rewrite this intro more formally?”
Use Case Example:
Indian Express uses ChatGPT in a conversational thread to generate multiple variations of a feature intro.
Best For:
Iterative refinement
Casual tone content
Drafting editorial pieces
Content idea generation
Tip: Treat the AI as a collaborator, not a machine. Provide feedback like: “Make this shorter” or “Use simpler terms.”
Zero-Shot Prompts
These prompts give no prior examples—only the task. You're relying on the AI's general training.
Structure:
“Explain quantum computing in simple language.”
“Write a video script for World Environment Day.”
Use Case Example:
BBC uses zero-shot prompting to test how LLMs handle unknown or new global events like a sudden coup.
Best For:
Quick tasks
First drafts
Spontaneous ideas
Limitation: May produce generic or inaccurate results if the context isn’t clear.
Few-Shot Prompts
These include a few examples before the actual task—helping the AI learn the desired format, tone, or depth.
Structure:
“Example 1: [Headline + Summary]
Example 2: [Headline + Summary]
Now write one for this: [New input]”
Use Case Example:
Reuters trains its in-house LLM to rewrite stock market stories by giving 3 example summaries before asking for a new one.
Best For:
Structured content (e.g., financial updates, weather, sports)
Repetitive tasks
Internal style replication
Tip: Use recent internal examples to make outputs newsroom-compliant.
3. Case Studies from Media Organizations
Let’s explore how major media organizations around the world use different prompt types in real workflows.
Reuters – Structured Informational Prompting
Use Case: Automated financial journalism.
Reuters feeds its AI with stock market data and then uses few-shot instructional prompts to generate earnings stories in minutes.
Prompt Example:
“Here are 3 examples of quarterly earnings reports. Based on this new data from Infosys, write a similar news story.”
Why it works: The consistency of tone and structure in financial reporting makes few-shot prompting highly effective.
Result: Speed and uniformity in financial content across time zones.
The New York Times – Editorial AI Experiments
Use Case: Creative feature planning and brainstorming.
NYT’s innovation team uses conversational creative prompting for brainstorming cover stories, visual experiments, and opinion content.
Prompt Example:
“Imagine a world where AI writes poetry about climate change. What would the first stanza look like?”
Why it works: Engages editorial teams in co-creation with AI.
Result: Unique angles for complex topics, used in prewriting and ideation.
Indian Express – Hybrid Prompting for Regional Coverage
Use Case: Drafting and translating regional news stories.
Indian Express employs instructional prompts for translating English news into Hindi, Malayalam, and Tamil using Gemini or ChatGPT.
Prompt Example:
“Translate this story to Malayalam, keeping journalistic tone and headline structure.”
Why it works: Saves time and effort for multilingual publications.
Result: Greater reach and accessibility across regional audiences.
The Hindu – Style-Adherent Prompting
Use Case: Tone refinement for news intros.
The Hindu uses role-based conversational prompting, asking AI to act like a Hindu subeditor and rephrase user-written intros.
Prompt Example:
“You are a senior editor at The Hindu. Rewrite this intro with a formal and factual tone.”
Why it works: Maintains traditional newsroom integrity while leveraging AI efficiency.
Result: Consistent editorial tone, especially in syndicated content.
BBC – Ethical AI Testing and Scriptwriting
Use Case: Testing hallucination rates, script generation for explainer videos.
BBC's R&D team uses a mix of zero-shot and few-shot prompts to evaluate how reliable AI outputs are for current affairs.
Prompt Example:
“Write a 1-minute explainer on the Ukraine war for a 15-year-old viewer.”
Why it works: BBC prioritises clarity and fairness in global reporting.
Result: Educational and age-appropriate content was made efficiently, backed by editorial verification.
4. Prompt Writing Tips Based on Type
Prompt Type Best Practices
Instructional Be precise with task, word count, tone, format
Conversational Allow follow-ups, give feedback, build context
Zero-Shot Make instructions detailed, avoid ambiguity
Few-Shot Use relevant examples, show tone & structure
Informational Ask for sources, summaries, cross-check facts
Creative Add storytelling tone, simulate scenarios
5. Integrating Prompt Types in the Media Workflow
Workflow Stage Suggested Prompt Type Example
Research Informational / Conversational “List recent events in Kashmir with dates.”
Drafting Instructional / Few-shot “Write a 300-word feature using these 3 samples.”
Visual Content Creative / Instructional “Generate a surreal Midjourney image about drought.”
Social Media Promotion Creative / Conversational “Give me 5 fun captions for this trending news story.”
Video Scripting Instructional / Zero-shot “Write a 90-second YouTube script about El Niño.”
Prompting Is Editorial Decision-Making
Prompts are not one-size-fits-all. In media, they serve different purposes—from structuring headlines and generating tweets to crafting metaphors and designing illustrations.
The best media professionals are those who learn to switch prompt modes based on task, audience, and platform. A deep understanding of informational vs creative goals, and the ability to choose between instructional, zero-shot, or few-shot prompts, can elevate both the quality and efficiency of content.
This isn’t just about getting better at AI. It’s about becoming a smarter communicator, a sharper editor, and a more adaptive media thinker.
As you continue this course, practice writing prompts in each category. Mix types. Experiment with tone. And remember: your prompt is the brief. The AI is the intern. You’re still the editor-in-chief.
Module 4: Writing Effective Prompts – Best Practices
Clarity, context, specificity
Role prompting and chain-of-thought prompting
Iteration and refining responses
Using temperature and tokens for control
Writing Effective Prompts – Best Practices
Prompt engineering has emerged as one of the most crucial skills in leveraging Generative AI, especially in media-related work. Whether you're a journalist, editor, social media strategist, or video producer, the effectiveness of your interaction with AI tools largely depends on how well you formulate prompts. A well-structured prompt can mean the difference between a bland, off-target AI output and a sharp, insightful, and creative response that meets editorial standards.
This module covers the essential best practices for writing effective prompts. From basic principles like clarity and specificity to advanced concepts like role prompting and adjusting parameters like temperature and tokens, this session will prepare you to master the art and science of prompting.
Clarity, Context, and Specificity
1. Clarity: Say What You Mean, Clearly
Clarity is foundational. AI models like GPT-4 interpret language based on patterns, probabilities, and context—but they do not "understand" like humans do. Vague, abstract, or ambiguous prompts often lead to generic or irrelevant results.
Example:
Unclear Prompt: “Write something about elections.”
Clear Prompt: “Write a 500-word news article about the voter turnout in the 2024 Indian General Elections, focusing on youth participation in Kerala.”
2. Context: Give the AI a Situation to Work With
Generative AI thrives on context. The more background information you provide, the better the model can tailor its responses. For media, this may include target audience, platform (e.g., social media, print, video), tone (e.g., formal, witty), and format (e.g., headline, report, infographic caption).
Example:
“You are an editor at a Malayalam daily newspaper. Rewrite the following news in a way that appeals to rural readers.”
Adding such framing can help generate content that feels far more human and situation-aware.
3. Specificity: Narrow Down the Target
Precision is key in media-related prompts. Instead of asking for “a post about AI,” ask for “a 280-character tweet about how AI is being used in Indian newsrooms.” The model performs better when it has a specific objective.
Role Prompting and Chain-of-Thought Prompting
1. Role Prompting: Assign an Identity to the AI
Role prompting makes the AI adopt a persona or expertise level. This significantly influences the tone, structure, and focus of the output.
Examples:
“You are a senior investigative journalist for the BBC. Summarize this press release into a hard-hitting news brief.”
“You are a fact-checking assistant. Identify possible factual errors in the given content.”
By defining a role, the AI can better align its responses with the expectations of that role.
2. Chain-of-Thought Prompting: Encourage Step-by-Step Reasoning
Chain-of-thought (CoT) prompting is a technique used to improve the reasoning of the AI. Instead of asking for a direct answer, you encourage the model to break down the problem into steps.
Example:
“Let’s think step-by-step. List the factors that led to the decline of newspaper sales in India, then explain each with an example.”
This is especially useful in data journalism, editorial strategy, or long-form narratives where layered analysis is required.
Advanced Chain-of-Thought Prompt:
"Imagine you're preparing an editorial piece on the ethics of deepfake technology. First list the pros and cons. Then expand on each with examples from recent global or Indian incidents. Finally, suggest how media houses can deal with it."
This guides the AI through a structured, logical path.
Iteration and Refining Responses
No matter how good your prompt is, the first response might not always be perfect. Iterative prompting allows you to improve results by progressively refining the prompt or the AI’s output.
1. Ask the AI to Improve Its Own Response
Example:
Initial prompt: “Write a headline for this article.”
Follow-up: “Make it shorter and more clickable for social media.”
2. Break Big Tasks into Smaller Prompts
Trying to get an AI to write a 2000-word report in one go? Often unwise. Instead:
First prompt: “List key points to include in the report.”
Second: “Expand on each point into paragraphs.”
Third: “Edit for journalistic tone.”
3. Prompt-Output-Refine Loop
Draft your initial prompt.
Analyse the AI’s response: Is it missing tone? Facts? Structure?
Re-prompt with adjusted instructions.
(Optional) Paste the output and say: “Rewrite this with a more analytical tone and include a headline.”
This feedback loop simulates a human editing process and ensures high-quality results.
Using Temperature and Tokens for Control
Generative AI platforms like ChatGPT allow you to set parameters for how the model behaves. Understanding “temperature” and “tokens” gives you more control over the output.
1. Temperature: Controls Creativity vs Precision
Low Temperature (0.1–0.3): More deterministic, factual, and conservative.
High Temperature (0.7–1.0): More random, creative, and free-flowing.
Use Case:
Low temperature: Writing legal disclaimers, fact sheets, or formal news.
High temperature: Brainstorming headlines, creative ads, or satire.
Prompt Example:
"Generate three headlines for this article. Set temperature to 0.8 for a more creative tone."
2. Tokens: Word Count & Budget
Tokens are chunks of words. For instance:
"The cat sat on the mat" = 6 tokens.
A 1000-word article might be around 1500 tokens.
Most AI models have a limit (e.g., 4096 or 8000 tokens per session). Long prompts or inputs can exhaust this, leading to truncated outputs.
Best Practice:
For long-form content, split into sections.
Use "continue" prompts to complete outputs when the model cuts off.
Examples of Effective Media Prompts
1. News Headline Generator:
"You're a sub-editor for The Hindu. Write 5 different headlines for this 300-word article about Kerala's new AI policy. Use a neutral tone."
2. Translation with Cultural Sensitivity:
"Translate this English paragraph into Malayalam, keeping the tone respectful and suitable for an older rural audience."
3. Data Storytelling:
"You are a data journalist. Interpret this Excel data (attached) and list 3 possible stories from the trends. Summarize each in 50 words."
4. Image Prompting for AI Tools:
"Generate a Midjourney prompt for an illustration showing a journalist surrounded by screens, overwhelmed with information, in a cyberpunk newsroom setting."
Prompt Failures: What to Avoid
1. Ambiguity:
“Write something for my blog.” → Too vague.
2. Overstuffing:
“Write an article, 800 words, emotional, fact-based, interview style, with SEO keywords, and include a tweet, and do it fast.” → Leads to poor results.
3. Lack of Role Clarity:
“Summarize this.” → What kind of summary? For whom? Why?
Fix:
“Summarize this for a 16-year-old student preparing for a civics exam.”
Best Practices in Summary
Principle Action
Clarity Use specific words and outcomes
Context Define audience, platform, tone
Specificity Narrow down task and content
Role Prompting Assign identities to the AI
Chain-of-Thought Encourage step-by-step reasoning
Iteration Refine and build on responses
Temperature Control creativity
Tokens Be aware of limits and chunking
Real-World Application: From Prompt to Publication
Scenario: You’re an editor at Indian Express, and you need a short explainer on AI in farming for Instagram.
Prompt:
“You’re a content creator for the Indian Express social media team. Write a 100-word explainer on how AI is used in Indian agriculture. Target: young urban readers. Format: Instagram carousel. Keep it snappy and friendly. Include a call-to-action.”
AI Response:
(Slide 1) 🌾 Did you know AI is revolutionizing farming in India?
(Slide 2) 🤖 From drone mapping to soil health monitoring…
(Slide 3) 🌱 Even predicting crop diseases before they hit!
(Slide 4) 📲 Follow us for more stories on AI & the future of farming!
This shows the power of a structured prompt guiding format, tone, audience, and even platform.
Writing effective prompts is not just a technical skill—it’s a creative editorial act. In the media landscape, prompt engineering is a force multiplier. It helps journalists write better, editors think faster, designers ideate visually, and communicators engage smarter.
By understanding clarity, context, specificity, role playing, step-wise reasoning, and tuning AI behavior with parameters, you can turn any AI tool into your creative ally. The future of digital storytelling will be shaped by those who not only tell good stories—but prompt them well.
Module 5: Prompting for Newsroom Workflows
Headlines, intros, summaries
Rewriting for tone and audience
Translation, transcription, fact checking
SEO and metadata generation
Hands-on labs with real assignments
Prompting for Newsroom Workflows
In the evolving landscape of journalism, newsrooms are increasingly integrating generative AI tools to streamline content creation, increase productivity, and reach wider audiences. Prompt engineering has emerged as a critical skill in this process. From writing headlines to verifying facts, the precision and creativity of prompts can directly influence the quality and effectiveness of news content. This module explores how prompting can revolutionise newsroom workflows by supporting every stage of content creation, refinement, and distribution.
Headlines, Intros, and Summaries: The First Impression
Headlines
Headlines are the most clicked-upon element in any news piece. A strong headline attracts attention, communicates urgency, and offers a promise of value. AI tools like ChatGPT, Gemini, and Claude can generate multiple headline variants from a single news draft using structured prompts.
Sample Prompt:
"Generate 5 clickable and concise headlines for a 500-word news report on the recent cyclone that hit the east coast of India. Keep the tone factual and suitable for a national daily."
AI can provide:
"Cyclone Ravages East Coast: Thousands Displaced"
"Monsoon Mayhem: Cyclone Devastates Andhra, Odisha"
The journalist can pick, modify, or combine suggestions, saving time and boosting creativity.
Intros
The intro or lead sets the stage. Prompting can help frame intros in different journalistic styles—straight news, narrative lead, question-based, or scene-setter.
Sample Prompt:
"Write an engaging intro paragraph (40 words max) for a story about rising onion prices in Maharashtra. Use a straight news style."
AI Output:
"Onion prices have surged to ₹80 per kilo in Maharashtra’s wholesale markets, sparking concern among consumers and traders ahead of the festive season."
Summaries
Summaries help readers scan and understand the gist quickly. Whether for newsletters, social media, or mobile alerts, AI-generated summaries are becoming the norm.
Sample Prompt:
"Summarize the following 800-word article into 3 bullet points for a mobile app notification."
This ensures editors can spend time on verification rather than rewriting.
Rewriting for Tone and Audience
A story written for a broadsheet may need adaptation for social media, TV, or youth-focused platforms. Prompting helps reshape the tone without changing the facts.
Example 1: From Formal to Conversational
Original: "The Finance Minister announced an economic relief package aimed at MSMEs."
Prompt: "Rewrite the sentence in a conversational tone for Gen Z readers on Instagram."
Output: "India just dropped a relief plan to help small businesses bounce back!"
Example 2: Rewriting for a Regional Audience
Prompt: "Rewrite the article in simple English for readers with limited financial literacy, focusing on how the new scheme helps farmers."
This function is particularly useful in multilingual or socio-economically diverse news markets like India.
Translation, Transcription, and Fact Checking
Translation
AI prompts can now handle cross-language translation with context preservation. Tools like Google Translate, ChatGPT (multilingual mode), and DeepL can assist.
Prompt Sample:
"Translate this 250-word news story from English to Malayalam, retaining journalistic tone and avoiding word-for-word translation."
Such prompting supports regional bureaus, vernacular publications, and hyperlocal journalism.
Transcription
AI tools like Otter.ai, Whisper, and Descript provide audio transcription. But for better accuracy, journalists can combine transcription with prompting.
Prompt Sample:
"Clean and format the following transcription for readability. Remove filler words and background noise text."
This is helpful for interviews, podcasts, and press briefings.
Fact Checking
Prompt-based fact-checking is an emerging use. While not foolproof, tools like ChatGPT and Gemini can identify inconsistencies or highlight claims needing verification.
Prompt Example:
"Does this statement: 'India is the world’s largest wheat exporter' align with current data as of August 2025? Cross-check and provide source suggestions."
Though human oversight remains essential, prompting accelerates initial verification.
SEO and Metadata Generation
Search Engine Optimisation (SEO) is essential for online discoverability. AI can generate keyword-optimised headlines, slugs, and tags when prompted correctly.
Prompt Example:
"Generate 10 SEO keywords for this article about women’s participation in Indian elections. Ensure high Google search relevance."
Metadata elements like description tags, canonical URLs, and image alt texts can also be generated.
Prompt:
"Write a meta description (150 characters) for an article titled: ‘Kerala’s Youth Revive Organic Farming.’ Include the phrase 'organic farming in Kerala'."
These practices enhance visibility without editorial teams having to manually optimize each story.
Hands-On Labs with Real Assignments
To reinforce these skills, hands-on labs are essential. Here are sample exercises:
🧪 Lab 1: Headline Generator Comparison
Task: Provide a 300-word news story to students.
Goal: Use ChatGPT and Gemini to generate five headlines each.
Deliverable: Compare tone, SEO strength, and relevance.
🧪 Lab 2: Multilingual Reporting
Task: Translate a news intro into Hindi, Malayalam, and Tamil.
Prompt: "Translate this into [language] with journalistic fluency and avoid colloquialisms."
Goal: Evaluate translation accuracy.
🧪 Lab 3: AI Rewrite Challenge
Task: Rework a formal news story for three platforms:
Broadsheet
Instagram reel script
YouTube news explainer
Deliverable: Submit rewrites and explain prompt logic.
🧪 Lab 4: Prompt for Fact Alert
Task: Use AI to find and flag potentially false or outdated data in a government press release.
Prompt: "Review the following data claims and suggest any that require citation or verification."
🧪 Lab 5: SEO Sprint
Task: Take a rough draft of a news article.
Prompts to be used:
Generate 10 SEO keywords
Write a meta description
Suggest an SEO-friendly slug
Goal: Publish-ready metadata package.
Integrating AI with CMS and Workflow Tools
Newsrooms use CMS (Content Management Systems) like WordPress, Drupal, or custom platforms. Prompted outputs can be integrated directly into these systems.
Example:
Use API-based tools like GPT or Claude to feed headline suggestions into editorial dashboards.
Use plugins that allow on-page prompting for copy edits.
Media organizations like BBC and Bloomberg are already building in-AI workflows. In India, The Indian Express uses AI to test headline impact, while The Hindu has experimented with AI-curated newsletters.
Balancing Speed with Accuracy
While AI and prompt engineering accelerate newsroom functions, there's a constant tension between speed and accuracy. Journalists must develop a mental checklist before accepting AI-generated outputs:
Is the tone correct?
Are there hallucinations?
Is the context accurate?
Is the content bias-free?
A prompt does not absolve responsibility—it is an extension of the journalist’s intent. Accuracy is ultimately a human responsibility.
AI as a Partner in Journalism
Prompting for newsroom workflows is not about replacing human journalists, but augmenting their abilities. Whether it's drafting a headline, refining a story, or generating multilingual content at scale, well-crafted prompts offer immense power when combined with editorial judgment.
This module teaches media professionals to:
Harness prompt engineering at various newsroom stages
Use AI creatively and ethically
Save time while maintaining journalistic standards
By the end of this session, learners will be proficient in designing, testing, and refining prompts tailored for journalism workflows—and confident in integrating them into real newsroom tasks.
Module 6: Prompt Engineering for Audio-Visual Content
Script generation for reels, videos, and podcasts
Prompting for visuals: DALL·E, Midjourney, Ideogram
Voice and dubbing tools (ElevenLabs, Descript)
Integrating prompts with editing software
Prompt Engineering for Audio-Visual Content
In today’s media landscape, where audiences consume more visual and auditory content than ever before, the role of artificial intelligence in shaping reels, videos, and podcasts is expanding rapidly. Visual storytelling has become central to journalism, brand communication, entertainment, and education. Prompt engineering is now a vital skill in producing high-quality, audience-relevant, and platform-specific audio-visual (AV) content.
This module is designed to help creators, journalists, and media students explore how to generate AV content using GenAI tools. From crafting video scripts to generating thumbnails, voiceovers, and prompts for visual art, learners will develop hands-on skills that can be directly applied in professional workflows. The module also highlights the tools—like DALL·E, Midjourney, Ideogram, ElevenLabs, and Descript—that are transforming content creation globally.
Script Generation for Reels, Videos, and Podcasts
Writing for audio and video is very different from writing for print or web. Scripts for reels, explainer videos, documentaries, or podcasts require a specific tone, flow, and emotional arc. With prompt engineering, creators can speed up this process without sacrificing creativity or intent.
1. Scriptwriting for Reels (Under 60 Seconds)
Reels need to hook the audience within the first 3 seconds and deliver a compelling message in under a minute. Prompts can guide AI to write these crisp, platform-ready scripts.
Prompt Example:
"Write a 45-second Instagram reel script on ‘Why coffee is addictive’. Tone: fun, youth-friendly. Include a hook, a quick science fact, and a snappy ending."
Output:
[Hook] “Ever wondered why your brain begs for coffee every morning?”
[Body] “Turns out, caffeine blocks adenosine, a chemical that makes you sleepy. Less adenosine = more alertness. But here’s the catch—it also triggers dopamine, the 'feel-good' hormone!”
[End] “So next time you sip your brew, know you’re literally sipping happiness!”
Learners can customize tone, style, and pacing with slight prompt tweaks.
2. Scriptwriting for Explainer Videos (1–5 mins)
Explainers need clarity, structure, and an educational tone. With AI, scripts can be quickly outlined and expanded.
Prompt Example:
"Create a 3-minute explainer video script on ‘How solar panels work’. Audience: high school students. Tone: informative and friendly."
The result can be reviewed, edited, and transformed into a storyboard.
3. Podcast Scripting and Outlining
Podcasts require not only scripts but also outlines for discussions and interviews. Prompt engineering helps hosts brainstorm themes, generate episode structures, and even simulate mock interviews.
Prompt Example:
"Create an outline for a 30-minute podcast episode on ‘The future of food in India’. Include segments, guest questions, and closing notes."
Prompting for Visuals: DALL·E, Midjourney, Ideogram
AI image generation has revolutionised the way visuals are created for media. Tools like DALL·E (OpenAI), Midjourney, and Ideogram offer artists and editors the power to create thumbnails, storyboards, infographics, and character designs through detailed text prompts.
1. Understanding Visual Prompting
Prompting for visuals involves describing:
Subject: What should be in the image
Style: Realistic, anime, oil painting, flat design, etc.
Mood/Tone: Bright, mysterious, dramatic
Framing/Angle: Close-up, bird’s eye view, wide shot
2. Using DALL·E for Editorial Visuals
DALL·E is ideal for editorial content due to its ability to generate accurate, coherent, and safe imagery.
Prompt Example:
"Generate an editorial-style illustration of a farmer in Kerala using solar-powered irrigation. Style: watercolor. Daylight scene."
This visual can be used as a cover image for an online story or newsletter.
3. Midjourney for Artistic and Conceptual Work
Midjourney is better suited for stylized, high-art, and surreal visuals. It requires highly descriptive prompts, and outputs are often more interpretive than literal.
Prompt Example:
"A futuristic newsroom run by AI, in cyberpunk style, with neon lights and holographic screens –– ultra-detailed, cinematic lighting."
Perfect for creating conceptual visuals for features and trend stories.
4. Ideogram for Typography and Posters
Ideogram is designed for text-in-image use—such as quotes, announcements, or digital posters.
Prompt Example:
"Design a digital poster with the text ‘Fact Matters’ –– bold letters, retro newspaper background, minimalistic colors."
Visuals like these are excellent for social media advocacy campaigns.
Voice and Dubbing Tools: ElevenLabs, Descript
Voice has become a cornerstone of content strategy—be it narration, dubbing, voiceovers, or audio translation. AI-powered voice tools are changing the way audio is produced across languages and formats.
1. ElevenLabs: Realistic Voice Generation
ElevenLabs is a leading tool in cloning voices and generating realistic voiceovers from text. It supports multilingual speech synthesis, custom voice creation, and emotional modulation.
Prompt Setup:
Upload script
Choose voice style (e.g., male/female, energetic/calm)
Select language
Use Case Example:
A Malayalam journalist uses ElevenLabs to create an English voiceover of their regional report for global audiences.
Prompt Tip:
"Read the following script in a warm, conversational tone suitable for a documentary narration. Voice should sound like a 35-year-old female from South India."
2. Descript: Audio Editing and Overdub
Descript allows you to edit podcasts and voice content like a Word document. Its Overdub feature lets you clone your voice, enabling quick fixes and multilingual dubbing.
Use Case:
Record an episode in English
Generate dubbed audio in Hindi using Descript and ElevenLabs
Edit errors via text instead of waveform manipulation
This accelerates content localization, especially for pan-Indian and global media houses.
Integrating Prompts with Editing Software
Prompt engineering doesn’t stop at generating raw material. It also integrates with video and audio editing software for smoother post-production.
1. Script-to-Storyboard Tools
Tools like Runway ML, Pika Labs, and Synthesia allow creators to input scripts and automatically generate storyboard frames, shot sequences, or even AI avatars delivering the lines.
Prompt Example:
"Generate a 5-shot storyboard for a 1-minute explainer video on 'How electric vehicles work'. Use flat animation style. Add annotations for camera angles."
The storyboard can then be used by editors in Premiere Pro or Final Cut Pro.
2. AI-Powered Video Editing Assistants
Many modern editors have plug-ins or native support for AI prompting.
Adobe Premiere Pro + AI Prompting
Use tools like Adobe Firefly to auto-generate background images
Generate captions via prompt: "Create social-media-friendly subtitles in a bold yellow font."
Final Cut Pro with Plugins
Prompt to auto-cut video based on script
Prompt to generate B-roll suggestions
3. Audio Cleanup and Sound Design
Prompting tools like Auphonic, Descript, and Krisp clean audio using natural language commands.
Prompt Example:
"Remove background noise and normalise the volume to podcast standards."
This helps journalists and creators ensure consistent sound quality, especially in field recordings.
Hands-On Labs and Exercises
To ensure practical understanding, this module includes lab-based assignments that allow students to test tools and craft their own prompts for AV content.
🎬 Lab 1: Reels Script Factory
Task: Write a 45-second reel script on a trending topic using ChatGPT
Deliverable: Generate 3 tone variants—funny, educational, sarcastic
Tool: ChatGPT or Claude
🎧 Lab 2: AI Voiceover Production
Task: Use ElevenLabs to create a voiceover of a written script
Deliverable: Submit voiceover in both English and one Indian language
Focus: Tone modulation and pronunciation accuracy
🎨 Lab 3: Visual Story Prompting
Task: Create 3 AI-generated images using Midjourney
Prompt: One photo-realistic, one artistic, and one abstract
Deliverable: A prompt-to-image mapping sheet
📼 Lab 4: Descript Audio Edit
Task: Edit a 3-minute podcast using Descript
Activities: Remove fillers, add intro/outro music, and insert voiceover
Deliverable: Before-and-after audio file
🎥 Lab 5: Video Prompt Integration
Task: Use Runway ML to animate a 4-line script
Goal: See how prompt inputs translate into visual motion
Deliverable: Animated short clip and reflection on prompt efficiency
Real-World Applications
Newsrooms
Generate explainer visuals for complex stories (e.g., elections, budgets)
Voice-dub videos for multilingual audiences
Auto-generate thumbnails and YouTube metadata
Independent Creators
Create content calendars by prompting daily reels topics
Clone own voice to automate narration
Use Ideogram to make social campaign art
Education and Training
Develop podcast-based learning content with scripted AI voices
Use storyboarding tools to teach media literacy
Translate educational videos using dubbing and subtitles
Ethical Considerations
With powerful tools come responsibilities:
Voice cloning must respect consent and copyright.
Visuals must not misrepresent reality in news contexts.
Deepfakes must be flagged or avoided in journalism.
Always:
Label AI-generated content clearly
Get consent for voice usage
Avoid misleading edits
Creating Visual Stories with Smart Prompts
Prompt engineering is not just a technical skill—it’s a creative force multiplier. For visual and audio content creators, mastering the art of prompting means unlocking entire worlds of rapid, high-quality content generation. Whether you are a budding journalist, an independent filmmaker, or a podcast producer, prompts are your new pen, brush, and mic.
By the end of this 6-hour module, learners will:
Be proficient in script prompting for different AV platforms
Generate visuals using leading GenAI tools
Use AI for multilingual voiceovers and podcast production
Seamlessly integrate prompts into popular editing workflows
The future of audio-visual storytelling is here—and it starts with a well-crafted prompt.
Module 7: Prompting for Investigative and Data Journalism
Data-to-narrative prompting
AI tools for scraping, parsing, and interpreting data
Prompting for interview simulation, analysis, and storytelling
Example: Panama Papers, Lok Sabha analytics, etc.
Prompting for Investigative and Data Journalism
In the modern era of journalism, investigative reporting and data journalism remain crucial tools for holding power to account, unearthing hidden truths, and creating impactful narratives. However, the methods and tools available to journalists have changed dramatically. As Artificial Intelligence (AI) becomes more accessible and sophisticated, prompt engineering is emerging as a powerful companion in the toolkit of investigative and data journalists. By crafting the right prompts, journalists can transform raw data into stories, automate tedious tasks, simulate interviews, and dive deeper into complex datasets.
This module focuses on how prompt engineering, powered by Generative AI and Large Language Models (LLMs), can be applied to the high-stakes world of investigative and data journalism. It explores the strategies, tools, and ethics involved, backed by practical examples like the Panama Papers, Lok Sabha analytics, and global case studies.
Data-to-Narrative Prompting: From Spreadsheets to Stories
At the heart of data journalism lies the challenge of transforming numbers, patterns, and trends into compelling human narratives. Traditionally, journalists relied on spreadsheets, charts, and human analysis to interpret datasets. Today, AI tools can assist in data-to-text generation — transforming raw information into readable summaries, headlines, and even long-form stories.
Prompt engineering makes this transformation possible. With the right prompt, an AI model like ChatGPT-4, Claude 3, or Gemini 1.5 can ingest a table or dataset and generate:
Summary of key findings
Trend analysis
Hypotheses for further investigation
Contextual storytelling
Example Prompt:
"Here is a CSV file of 2024 Lok Sabha election results by constituency. Identify the 10 constituencies with the narrowest winning margin and suggest why these areas may need closer political analysis."
This kind of prompt allows the journalist to focus on editorial decision-making, while the AI handles the data crunching.
Use Cases:
Voter turnout patterns over decades
Crime data visualisation and summary
Budget analysis of government schemes
Misinformation trends based on social media metrics
Prompt engineering allows for multimodal prompts too. Journalists can combine text, numbers, charts, and even images to produce richer insights.
AI Tools for Scraping, Parsing, and Interpreting Data
Investigative journalists often spend hours scraping web data, decoding PDF files, and manually sorting information. AI-powered tools now enable automated scraping and interpretation through prompt-based interactions.
⚙️ Useful Tools and Their Prompting Capabilities:
BeautifulSoup + GPT:
Scrapes structured HTML data and summarizes or analyzes content.
Prompt: “Scrape all press releases mentioning land acquisition from this government website and summarize their content in 100 words each.”
ChatGPT Code Interpreter / Advanced Data Analysis (ADA):
Handles CSVs, Excel files, and performs statistical analysis with simple language.
Prompt: “Analyze this CSV of sanitation-related budget allocations by Indian states from 2015 to 2025. Highlight any major shifts in spending patterns.”
Diffbot / ParseHub / Octoparse:
Automate web crawling and integrate with LLMs for further interpretation.
Prompt: “Extract the list of companies involved in offshore banking from this set of leaked PDFs and match them with Indian registered business names.”
Tabula + LLM:
Extract tables from scanned PDFs and process them for insights.
Prompt: “Summarize financial irregularities from the 2021 report of the Comptroller and Auditor General of India (CAG). Focus on discrepancies in procurement.”
With tools like ChatGPT’s file upload, a journalist can drag and drop a scanned government order, then use a carefully crafted prompt to extract core information in seconds.
Prompting for Interview Simulation, Analysis, and Storytelling
AI models can be used not only to summarize or interpret data, but also to simulate conversations — a revolutionary capability for both investigative leads and pre-interview preparation.
Interview Simulation
Let’s say a journalist is preparing to interview an official about discrepancies in housing scheme allocations. The AI can be prompted to simulate possible responses based on public data, speeches, or historical patterns.
Example Prompt:
"Assume the role of a district officer justifying why PMAY housing allotments in rural Kerala have reduced since 2020. Simulate a conversation with a journalist asking about delays, fund allocation, and selection process."
This helps the journalist frame sharper real-life questions, anticipate evasive answers, and identify areas requiring deeper scrutiny.
Content Analysis
Once the actual interview or press conference is done, the AI can assist in:
Transcription (Descript, Whisper)
Sentiment analysis
Fact-checking and contradiction detection
Creating summaries or narrative arcs
Story Generation
Given raw interview transcripts and related data, a well-crafted prompt can help generate compelling story drafts.
"Using this interview with a whistleblower and data from the audit report, create a 1000-word narrative exposé in the style of The Wire, focusing on the misuse of rural development funds."
This doesn’t replace the journalist but enhances creativity and saves time, allowing more energy for fact-checking, fieldwork, and ethics.
Case Studies in Prompting for Investigative Journalism
1. Panama Papers (2016) – ICIJ
The Panama Papers leak involved over 11.5 million documents revealing how wealthy individuals used offshore tax havens. Traditional journalistic methods would not have scaled.
Had today's AI tools been available, journalists could have used:
Prompt-based parsing of documents in multiple languages.
Entity extraction via LLMs (e.g., “Find all Indian citizens mentioned in these files.”)
Visual mapping tools to show relationships.
Prompt engineering would allow collaborative, cross-border investigative work, with unified instruction sets for various AI agents.
2. Lok Sabha Analytics – IndiaSpend / The Hindu
With datasets on parliamentary attendance, question types, and constituency development fund usage, AI models can generate story leads.
Prompt: “From this dataset of questions raised in Parliament, identify MPs who have focused on environmental issues. Rank them by frequency and state representation.”
Such prompts help find underreported trends, emerging voices, and regional patterns — often the first step toward bigger investigations.
3. BBC’s Disinformation Unit
BBC journalists used AI to monitor fake news trends during elections. With smart prompting, AI models were trained to:
Identify deepfake videos
Summarize viral posts on Telegram or Facebook
Cluster similar misinformation narratives
Prompt:
“Analyze this set of viral images circulating on WhatsApp. Determine which may be AI-generated and provide reasons.”
This kind of rapid response journalism relies on prompt fluency, accuracy, and ethical awareness.
Ethical Considerations in Investigative Prompting
Investigative journalism holds immense public value — and with it comes ethical responsibility. Prompt engineering must be designed to amplify truth, not distort it.
Key Ethical Guidelines:
Transparency: Declare AI use in the reporting process.
Verification: AI-generated hypotheses or claims must be fact-checked by humans.
Bias Awareness: LLMs are trained on biased data; prompts should account for that.
Consent & Privacy: Simulated interviews must never be presented as real.
Avoid Automation Bias: Don’t rely on AI-generated results without editorial judgment.
Training Journalists: Hands-on Prompt Labs
Here are practical lab assignments for students and journalists:
🧪 Lab 1: From PDF to Story
Input: A 100-page CAG report on health funding.
Task: Extract irregularities using LLM + Tabula + ChatGPT.
Prompt: “Summarize five financial anomalies in Kerala’s health department funding between 2018 and 2023.”
🧪 Lab 2: Election Data Analysis
Input: Lok Sabha 2024 results CSV + voter turnout history.
Prompt: “List 10 constituencies with the highest change in vote share. Provide possible reasons and link to major local issues.”
🧪 Lab 3: Simulated Interview + Writeup
Prompt: “Roleplay an RTI activist who uncovered irregularities in MGNREGA payments in Bihar. Respond to questions by a journalist investigating further.”
🧪 Lab 4: Social Media Misinformation Tracker
Input: Screenshots from viral posts on Telegram, X (formerly Twitter).
Prompt: “Group these messages by misinformation type. Are there common visuals or language being used?”
These exercises train journalists to use AI not just as a tool, but as a thinking partner.
AI as a Watchdog’s Ally
Prompt engineering is not about automation for its own sake. In the hands of investigative and data journalists, it becomes a force multiplier — allowing deeper dives, faster turnarounds, and broader access to patterns and insights. By learning to communicate effectively with machines, journalists can preserve the watchdog role of the press in the digital age.
This module encourages learners to see prompting as part of the investigative mindset: curious, skeptical, strategic, and always rooted in truth. With the rise of AI, the future of journalism is not about man versus machine — it’s about journalist plus machine, working together for public good.
Module 8: Prompting for Social Media and Audience Engagement
Platform-specific content (X, Instagram, Threads, YouTube)
Engaging titles, captions, hashtags
Meme and trend generation with AI tools
Prompting for comments and community management
Prompting for Social Media and Audience Engagement
In today's rapidly evolving media landscape, social media has become the frontline of news delivery, audience engagement, and brand building. For media professionals, this means staying relevant not just through content creation, but by crafting content that adapts to platforms, reacts to trends, and connects with communities in real time. Enter prompt engineering—the powerful skill of guiding AI models to create social-first content that is fast, relevant, engaging, and platform-aware.
Prompt engineering isn’t just a tech-savvy shortcut anymore. It’s emerging as a critical skill in the social media playbook, helping media houses, content creators, and journalists customize their voice across platforms like X (formerly Twitter), Instagram, Threads, and YouTube. With the right prompts, you can generate high-performing posts, adapt content for specific audiences, create memes, identify hashtags, moderate comments, and even mimic brand tone.
This module covers how to prompt effectively for social platforms using generative AI, with deep dives into platform-specific styles, AI-driven trend analysis, audience interaction, and more.
Platform-Specific Content Creation
Each social media platform has a unique content language, algorithm, user expectation, and rhythm. Generative AI can help media professionals tailor content based on these nuances by feeding it with precise, style-aware prompts.
1. X (formerly Twitter)
Nature: Text-dominant, high-speed news cycles, virality-driven
Prompt example: “Write a breaking news tweet about a train derailment in Kerala in under 280 characters with a link and a sense of urgency.”
What AI can generate: Crisp headlines, thread ideas, quote tweets, and poll formats
AI tip: Use zero-shot prompts for instant tweets, and chain prompts for developing threads with layered information
2. Instagram
Nature: Visual-first, aesthetic-driven, captions that connect
Prompt example: “Write a poetic caption for a sunset over Fort Kochi with local cultural references and 5 trending hashtags.”
What AI can generate: Story ideas, carousel text, Reels captions, CTA in bios
AI tip: Include tone (e.g., witty, nostalgic, poetic) and audience (e.g., Gen Z, travelers)
3. Threads
Nature: Meta’s conversational alternative to X, still evolving
Prompt example: “Draft a conversational, non-toxic, culturally aware thread on climate change effects in Delhi, in 3 posts.”
What AI can generate: Friendly tone content, quote response drafts, daily prompts
AI tip: Prompt for community-safe and engaging content by adding “sensitive to current discourse” in the prompt
4. YouTube
Nature: Long-form video, but titles, descriptions, tags matter greatly
Prompt example: “Write a compelling title and description for a 10-minute video about India's moon mission Chandrayaan-3 for a young audience.”
What AI can generate: SEO-optimised titles, video summaries, script hooks, thumbnail text
AI tip: Prompt with “optimise for YouTube algorithm and high CTR”
Engaging Titles, Captions, and Hashtags
The first impression of social media content often lies in its title or caption. This is what grabs attention, communicates tone, and drives engagement.
1. Headlines to Hook
Headlines on social media aren’t just about summarizing—they must provoke curiosity.
Prompt: “Create 5 click-worthy titles for a post about India’s plastic ban, in different tones: informative, humorous, urgent, emotional, and ironic.”
2. Caption Crafting
Captions on Instagram or YouTube aren’t throwaway texts. They reflect identity, creativity, and brand.
Prompt: “Write an Instagram caption in Hinglish that blends humor and information for a photo showing water scarcity in Bangalore.”
3. Hashtag Research & Generation
Hashtags help discoverability. AI can generate based on context, region, and niche.
Prompt: “Suggest 10 trending hashtags for a video about Onam festival in Kerala with a mix of local and global appeal.”
You can also prompt AI to analyze trending hashtags using tools like Google Trends, KeywordTool.io, or Hashtagify.
Meme and Trend Generation with AI Tools
Memes are one of the fastest-growing and most engaging content formats. Prompt engineering for memes means combining humor, format familiarity, and topical relevance.
1. Text-based Meme Prompts
Prompt: “Write a meme caption in the ‘distracted boyfriend’ format, representing a media outlet distracted by AI trends while ignoring journalistic ethics.”
2. Visual Meme Generation
Use tools like DALL·E, Midjourney, Ideogram to create visual memes through prompts.
Prompt: “Create a cartoonish image of an exhausted journalist sitting in front of five AI tools all asking for prompts, in 16:9 format.”
3. Reaction & Satire Memes
Prompt: “Write a sarcastic meme text about influencer apologies using the ‘apology notes’ trend.”
Prompting for memes requires cultural awareness and sensitivity. AI may miss nuance, so always verify tone and implications.
Prompting for Comments and Community Management
Maintaining a social presence goes beyond posting—it requires interaction, moderation, and engagement. AI can help draft responses, clarify user questions, and even manage trolls with diplomacy.
1. Writing Polite Responses
Prompt: “Reply diplomatically to a user who criticizes our news report on a political rally, while explaining our editorial policy.”
2. Comment Thread Moderation
AI can help summarize heated conversations and suggest moderation steps.
Prompt: “Summarise this 50-comment thread and suggest a community guideline reminder post that can de-escalate.”
3. Creating Community Prompts
You can use AI to generate prompts that spark engagement:
Prompt: “Create 5 open-ended questions we can post on Threads to engage readers on the topic of AI in education.”
AI Tools for Social Media Prompting
Here are some powerful tools media professionals can use in tandem with prompt engineering:
Tool Use Case
ChatGPT / Claude Caption, headline, idea generation
Copy.ai / Jasper Platform-specific post creation
Hashtagify Trending hashtag insights
Midjourney / DALL·E Visual meme and concept art
RunwayML / Pika / Kaiber Video transformation and reels
ElevenLabs / Descript Voiceovers, social podcast clips
Hootsuite / Buffer AI Assistants Scheduled post writing and optimization
Each tool performs better with specific types of prompts. For example:
ChatGPT excels in linguistic nuance
Jasper is designed for marketing content
Midjourney requires image-formatting prompts with camera angles, style tags, etc.
Best Practices in Social Media Prompt Engineering
1. Always specify the platform
Every platform’s audience is different. Begin your prompt with: “Write a post for Instagram...” or “Generate a tweet thread...”
2. Use defined tones
Add: “in a sarcastic tone,” “empathetic,” “authoritative,” or “Gen Z slang” to shape voice.
3. Mention character/word limits
Most platforms have restrictions. Example: “In under 280 characters.”
4. Add cultural/geographic context
AI defaults to a Western audience. Say “for Indian readers in Tier 2 cities” or “for Malayalam-speaking Gen Z.”
5. Test variations
One prompt can give multiple outputs. Generate 5–10 and A/B test manually or using platform analytics.
Mini Case Studies
1. BBC Hindi on Instagram
Uses poetic language in captions with emojis, and descriptive visual Reels.
Likely prompt: “Summarize this news about farmers’ protest in poetic Hindi under 150 words with emotional tone.”
2. Reuters on X
Short, sharp, factual tweets with minimal adjectives.
Likely prompt: “Write a breaking tweet about Turkish elections with factual tone and Reuters style guide in mind.”
3. The Hindu’s Reels
Educational yet trendy, often using pop culture.
Likely prompt: “Create a reel script combining Chandrayaan-3’s landing with a Marvel reference to engage youth.”
Exercises and Labs for Learners
Prompt Challenge:
Write 5 captions for the same photo of a climate rally—one for each platform (X, Instagram, Threads, YouTube, LinkedIn).
Visual Prompting Lab:
Use Midjourney or Ideogram to create a meme poster using a specific AI prompt.
Response Writing Drill:
AI generates comments on a sensitive post. Learners must write professional, human-like replies using ChatGPT.
Trend Prompting:
Use a trending news story to create a meme, reel caption, and thread using different tones—satirical, informative, and nostalgic.
In the noisy, scroll-driven world of social media, prompting for relevance is just as important as reporting for accuracy. With the right prompt engineering, media professionals can not only create content faster but also tailor it to the personality of each platform, tap into real-time trends, and nurture authentic conversations with their audience.
Social media is no longer just a place for distribution—it’s where media content lives, evolves, and breathes. Prompt engineering is the new grammar of that space. By mastering it, journalists and media creators unlock the potential to be more creative, responsive, and impactful than ever before.
Module 9: Legal, Ethical, and Accuracy Issues
Copyright and authorship
AI hallucination and misinformation
Bias in AI tools
Media ethics and editorial responsibility
Legal, Ethical, and Accuracy Issues in Prompt Engineering for Media
The rise of generative AI has introduced powerful tools for journalism, storytelling, and content creation. However, with great power comes immense responsibility. As media professionals increasingly integrate AI-driven outputs into their workflows, it becomes critical to understand the legal, ethical, and accuracy-related challenges that accompany the use of AI. Prompt engineering—designing queries and instructions to get useful responses from AI—has emerged as a core skill in this space, but its use must be guided by legal literacy, ethical clarity, and a firm grounding in journalistic responsibility.
This module addresses four core areas of concern:
Copyright and authorship
AI hallucination and misinformation
Bias in AI tools
Media ethics and editorial responsibility
Together, these topics equip journalists and content creators with the awareness to navigate the grey zones of AI-driven content creation.
1. Copyright and Authorship in AI-Generated Media
Who owns AI-generated content?
The question of authorship in AI-generated content is still evolving legally around the world. When a journalist uses a tool like ChatGPT or DALL·E to generate text or images, who owns the result? Is it the human who wrote the prompt? The company that built the AI? Or does it fall into the public domain?
In most jurisdictions today, AI-generated content without meaningful human intervention is not eligible for copyright. For example, the United States Copyright Office has ruled that content created entirely by AI lacks human authorship and therefore cannot be copyrighted. However, when a journalist crafts a detailed, creative prompt and edits the result meaningfully, they may assert copyright over the final product.
Copyright risks when using AI
AI models are trained on massive datasets, including books, articles, photographs, and artworks. In many cases, the data used for training is copyrighted material scraped from the internet. This opens up legal risks when journalists republish or monetize content generated by these models.
For instance:
A photo generated by DALL·E might accidentally mimic the style of a copyrighted image.
A news summary generated by an LLM might too closely paraphrase an original source.
Some major lawsuits have already emerged. Authors like Sarah Silverman and organizations like Getty Images have sued AI companies for using copyrighted content in training without permission. This puts both developers and users of AI at legal risk if they are not cautious.
Best practices:
Use AI tools that offer transparency about training data and copyright licensing.
Always edit and fact-check AI outputs before publication.
Avoid using generated content as-is unless explicitly licensed for commercial reuse.
2. AI Hallucination and Misinformation
What are AI hallucinations?
AI hallucinations occur when a language model like ChatGPT produces content that is grammatically correct and confident—but factually wrong. These errors aren't bugs; they're a consequence of how LLMs are trained. They're not connected to real-time databases unless explicitly designed to be.
For example:
An AI might invent quotes or misattribute statements.
It may list incorrect statistics with convincing confidence.
It can fabricate news stories, interview excerpts, or even references to fictional documents.
In a newsroom setting, this is dangerous. If a journalist relies too heavily on AI outputs without cross-checking, they risk publishing false or misleading information.
Risks of misinformation amplification
AI can also unintentionally amplify fake news, conspiracy theories, or biased narratives embedded in its training data. If media professionals use AI to "repackage" existing misinformation, it lends false credibility to those narratives.
Consider this:
Asking an AI to summarize a fake viral news story might give it a polished, trustworthy tone.
Using AI for translation or rewriting may remove context that reveals misinformation.
Best practices:
Always fact-check AI-generated outputs.
Use AI as a tool for research, not a substitute for journalistic verification.
Label AI-generated content clearly in your newsroom workflow.
3. Bias in AI Tools
How AI learns bias
AI tools reflect the data they are trained on. If the data includes biases—racial, gender, cultural, ideological—the AI will likely reproduce or even amplify those biases.
Examples in media settings:
AI tools may describe political leaders differently based on their gender or nationality.
Image generators may show stereotypical portrayals of professions (e.g., doctors as men, nurses as women).
Translations or summaries may reflect biased interpretations of news events.
Case studies:
In 2023, researchers found that image-generation tools underrepresented Black women in professional roles.
Some AI chatbots mirrored Western biases when describing conflicts in the Middle East or Africa.
Media organizations have discovered that tools like ChatGPT may censor or favor specific viewpoints when asked politically sensitive questions.
Bias is not just a technical issue—it's a journalistic one. If unchecked, it erodes public trust.
Best practices:
Be aware of how your prompts may lead to biased responses ("prompt bias").
Diversify the sources you use to validate AI-generated content.
Involve diverse editorial teams when interpreting or publishing AI-driven content.
4. Media Ethics and Editorial Responsibility
AI cannot replace editorial judgment
Journalism is not just about information; it’s about responsibility. Even as AI becomes a powerful assistant, journalists must maintain editorial oversight. Automation cannot excuse lapses in accuracy, fairness, or accountability.
Key ethical concerns:
Transparency: If AI was used to generate or assist content, should that be disclosed to the audience?
Accountability: If an AI-generated article spreads false information, who is held responsible?
Consent: Is it ethical to use AI to simulate interviews with public figures who never gave consent?
These questions don’t have universal answers yet, but each media organization must adopt clear editorial guidelines.
Emerging media standards
Leading global media organizations are already setting frameworks:
The New York Times forbids publishing AI-generated content without significant human editorial input.
BBC includes a disclaimer when AI tools are used to assist reporting.
Reuters has developed an internal AI Ethics Charter emphasizing transparency, accuracy, and human oversight.
In India, The Hindu and Indian Express have begun integrating AI for headlines and SEO but maintain strict editorial control over final stories.
Developing your newsroom code of conduct
A best practice for media organizations using AI is to create their own AI Editorial Code, which might include:
What tasks can AI assist with (summarizing, translation, etc.)
What content must remain human-generated (opinion, sensitive reporting)
When and how to disclose AI involvement
Rules for accuracy and fact-checking before publishing
Ethical Prompting Techniques
Prompt engineers in journalism must also act ethically while designing prompts. Prompting strategies can influence outcomes, tone, and even ideological leaning.
Examples:
Prompt: “Explain why X policy is failing” vs “Provide pros and cons of X policy”
→ The first prompt introduces bias from the start.
Prompt: “Write a glowing tribute to this political leader”
→ Not appropriate for objective reporting.
Prompt: “Summarize the speech of the Prime Minister from a neutral perspective”
→ More ethical and objective.
Being mindful of prompt phrasing is essential in preventing skewed or manipulative outputs.
Legal Frameworks in Progress
Governments around the world are moving towards regulating AI, but legislation is still catching up with technology.
Key developments:
European Union AI Act (2024): Requires transparency, traceability, and human oversight for high-risk AI systems, including those used in media.
India's DPDP Act (Digital Personal Data Protection Act, 2023): Focuses on protecting individual data rights, which can impact how training data for AI is used.
US AI Bill of Rights (proposed): Emphasizes the right to clear explanations and freedom from algorithmic discrimination.
As laws evolve, media professionals must stay updated and proactive in compliance.
Prompt engineering has opened up powerful possibilities for media professionals. From drafting news stories to creating visuals and videos, AI tools are rapidly transforming how journalism is practiced. But this power comes with new forms of risk—legal, ethical, and editorial.
Understanding and addressing these risks is not optional. It is foundational to responsible journalism in the age of generative AI.
A well-crafted prompt may create content in seconds, but only a thoughtful, ethically grounded journalist can ensure that the content meets the standards of truth, fairness, and public trust. As AI becomes more integrated into newsrooms, the responsibility to uphold journalistic integrity becomes more important—not less.
In short, AI can write, but only humans can be accountable.
Module 10: Capstone Project & Evaluation
Choose one: Long-form news story, podcast, YouTube explainer, or social media campaign – generated with AI
Peer review and feedback
Viva and reflective presentation
Certification
Final Reflections
At its heart, journalism is about telling the truth, creatively and courageously. Prompt engineering does not change that — it enhances our capacity to do so with new tools.
But tools are only as ethical as their users.
As media professionals, our mission is to use AI wisely, question it fiercely, and deploy it for good. Prompt engineering will help us ask better questions — not just of AI, but of power, policy, society, and ourselves.
So write prompts that imagine better worlds. That amplify unheard voices. That explore nuance. That challenge assumptions. That serve public interest. That open conversations, not close them.
Let your prompts become the starting point for stories that matter.
The future of journalism isn’t machine-made. It’s human-guided, AI-augmented, and mission-driven.
Congratulations on completing this journey. Now go, and prompt the change you wish to see. If keen on a Certificate please write to schoolofexecutiveeducation@gmail.com
Credits:
School of Executive Education