ChatGPT Prompts for Case Studies: 25 Templates

The short version: These 25 ChatGPT prompts for case studies cover interview prep, structure, drafting, headlines and repurposing, and every single one of them only works if you feed the model real client data first. The uncomfortable bit nobody tells you: left alone, ChatGPT will cheerfully invent a "47% increase in qualified leads" that never happened, so your job is to control the inputs, not just copy the outputs. Do that and you can go from raw interview notes to a publish-ready case study in about 90 minutes instead of a full day.

Why I still write case studies by hand-ish, sixteen years in

I've written or edited well over 200 case studies since I started consulting. Not one of the good ones came from typing "write me a case study about my client" into a chat box and hitting enter. The good ones came from a messy 45-minute Zoom call, a transcript full of "um" and half-finished sentences, and then me and ChatGPT working through that transcript together to find the actual story hiding inside it.

Last year I did exactly this for a UK cybersecurity SaaS client. The founder rambled for 45 minutes about implementation headaches, a support ticket backlog, and one specific Tuesday when a customer nearly walked. There was no headline in that transcript. There was a story. I ran the raw transcript through three of the prompts below, pulled out the numbers he'd mentioned (ticket resolution time dropped from 3 days to 6 hours, not the vague "faster support" he wanted to write), and the finished case study went live on his site three weeks later. It generated 14 inbound demo requests in its first month and became his most-shared LinkedIn post of the quarter. That result came from specificity, not from AI polish.

The rule that makes every prompt below work

Here's the part most "AI prompts for case studies" posts skip: ChatGPT has no idea what your client's results were. If you ask it to write a case study and don't paste in a transcript, a survey response, or at least three real bullet points from the client, it will fill the gaps with plausible-sounding fiction. I've seen agencies publish case studies with a made-up "62% ROI increase" because the AI needed a number and nobody checked. That's not just embarrassing, in the UK it can fall foul of ASA rules on substantiated claims if the case study is used in ads.

So the workflow is always: gather real material first (call recording, transcript, email thread, survey answers, analytics screenshot), then paste that material into the prompt along with the instruction. Every prompt below assumes you're doing that. If you skip it, you're not writing a case study, you're writing marketing fiction with a client's name on it.

How to run these templates

• Record or transcribe the client interview (Otter, Fireflies, or even your phone's voice memo app transcribed through ChatGPT itself)
• Paste the raw transcript into the prompt, not a summary you wrote from memory
• Run the structure prompts first, then the drafting prompts section by section
• Fact-check every number against the client's own dashboard or their written confirmation before publishing
• Send the client the final draft for sign-off, always, no exceptions

If you want somewhere to keep these so you're not hunting through old chat threads every time, I wrote a full walkthrough on building a personal library of ChatGPT prompts that's worth doing before you touch any of the templates below.

5 prompts for the interview and discovery stage

These come before you write a word. Skip this stage and every case study you produce will sound the same, because you'll be filling in gaps with generic language instead of the client's own words.

• "I'm interviewing a client called [name/company] about [product/service]. Write me 12 open-ended interview questions that will surface specific numbers, a turning-point moment, and at least one thing that didn't go smoothly."
• "Here's my client's industry: [industry]. What are the three metrics their buyers care about, that I should make sure I ask about in the interview?"
• "Read this raw transcript [paste transcript]. Pull out every number, date, and specific detail mentioned, even small ones, and list them separately from the general commentary."
• "Based on this transcript [paste transcript], what's missing that I should follow up with the client about before I write anything? Be specific about which claims have no number attached."
• "Turn these five bullet points from a client survey into three follow-up questions I can send by email to get more specific detail."

5 prompts for structuring the case study

Structure is where AI earns its keep, because a case study skeleton is a solved problem and you don't need to reinvent it every time.

• "Using the classic Problem, Solution, Results structure, build me a section-by-section outline for a case study using this transcript [paste transcript]. Flag which sections currently have no supporting data."
• "Suggest a structure for a case study that leads with the result in the first two sentences rather than the background, using this material [paste notes]."
• "I have a B2B SaaS client story with a 6-month timeline. Give me an outline that shows before/during/after rather than problem/solution/result, and tell me which format suits this content better."
• "Build a one-page executive summary structure for this case study that a busy CFO could read in 90 seconds, using only the strongest three data points from this transcript [paste transcript]."
• "Turn this outline into a bullet-point structure suitable for a sales deck slide rather than a web page, keeping it under 40 words per slide."

5 prompts for drafting the actual sections

This is where you feed sections in one at a time rather than asking for the whole thing at once. One long prompt for a full case study almost always produces bland, evenly-weighted prose where every sentence carries the same amount of importance. Drafting section by section gives you room to push back.

• "Write the 'Challenge' section of this case study using only these direct quotes and facts from the client [paste relevant transcript section]. Keep it under 150 words and use the client's own phrasing where possible rather than corporate language."
• "Write the 'Solution' section describing how we implemented [product/service], using this timeline [paste timeline]. Avoid words like '' and 'streamlined', they've been used in every case study on the internet."
• "Write the 'Results' section using only these confirmed numbers [list numbers]. Do not add any additional statistic, estimate, or percentage that isn't in this list."
• "Rewrite this results paragraph so it leads with the number instead of the sentence structure it's currently in, and cut anything that reads like an assumption rather than a fact."
• "Write a short client quote in the voice of someone from this transcript [paste transcript section], staying as close as possible to their actual wording, then show me the original line it's based on side by side so I can check it's not putting words in their mouth."

3 prompts for headlines and hooks

The headline decides whether anyone reads section one. Most case study headlines are dead on arrival because they name the company instead of the result.

• "Give me 10 headline options for this case study, each one leading with the specific number or outcome rather than the client's company name."
• "These are my current draft headlines [list them]. Rank them by how specific they are and tell me honestly which ones sound like every other SaaS case study on the internet."
• "Write a one-sentence hook I could use in the meta description and the opening line of the case study, using this confirmed result [paste result]."

4 prompts for formats and repurposing

One good case study should feed a dozen pieces of content, not sit on a single web page collecting dust. This is where I move between tools depending on the job. I break down exactly when I'd pick each one in this guide to ChatGPT vs Claude vs Gemini for marketing, but for repurposing case studies specifically, ChatGPT and Claude both handle it well as long as you feed them the finished, fact-checked draft rather than the raw transcript again.

• "Turn this finished case study [paste it] into a 200-word LinkedIn post written in first person from the client's perspective, not the company's."
• "Turn this case study into a 60-second video script with a hook in the first five seconds and a specific number by the fifteen-second mark."
• "Write three short subject lines and a 90-word email based on this case study that I could send to a segment of prospects in a similar industry to the client."
• "Suggest three visual concepts (a before/after chart, a stat callout, a quote card) I could brief for this case study, including exactly what text should appear on each one." If you're producing those visuals yourself rather than briefing a designer, my walkthrough on using an AI image generator for marketing covers how to get clean, on-brand results without the six-fingered-hand problem everyone jokes about.

3 prompts for editing and quality control

This is the stage people skip, and it's the one that protects your reputation.

• "Read this case study draft and flag any sentence that states a specific number, percentage, or timeframe. List them all in one place so I can verify each one against source material."
• "Read this draft and tell me honestly which sentences sound generic enough that they could apply to any client in any industry. Be blunt."
• "Compare this draft against the original transcript [paste both] and flag anywhere the draft adds a claim, detail, or number that isn't in the transcript."

Where this all breaks down if you're not careful

I'll say the uncomfortable part plainly because I think most posts on this topic dance around it: the moment you let ChatGPT write results you haven't verified, you've stopped writing a case study and started writing copy that happens to have a real client's logo on it. I've reviewed drafts from other consultants where the "before" numbers were softened and the "after" numbers were rounded up, not maliciously, just because the AI smoothed things over and nobody checked. That's how a good result (say, support tickets dropping from 3 days to 6 hours) quietly becomes an invented one ("resolved 90% faster") that a sharp-eyed prospect, or worse a journalist, could pick apart.

The fix isn't complicated, it's just tedious: every number in the final draft gets traced back to a source before it goes live. Not "the client seemed happy about it", the actual dashboard screenshot or the actual line in the transcript. It adds maybe 15 minutes to the process. It's the difference between a case study you can defend and one you'd rather nobody scrutinised.

Making this part of a repeatable system, not a one-off scramble

If you're producing case studies regularly, this whole process is a good candidate for automation, not the writing itself, but the boring bits around it: transcribing calls, pulling numbers into a spreadsheet, drafting the follow-up email asking the client for sign-off. I cover how I've set up similar workflows in how to use ChatGPT agents to save 10+ hours a week, and case study production is one of the clearer wins because the steps repeat almost identically every time. If you'd rather someone build that system for you than piece it together yourself, that's exactly the kind of thing an AI consultant for small business should be setting up in your first month working together, not your twelfth.

And if you want the writing itself to stop sounding like every other case study on your competitor's site, the deeper fix is in your source material, not your prompts. I go into this in how to use ChatGPT for content creation without sounding like everyone else, but the short version is: the transcript is where the personality lives, the prompt just gets it onto the page in the right order.

I keep every related walkthrough in the ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps. To write your own, try the free tool that writes the prompt for you.

Related: cybersecurity: guidelines and how to pitch.

Frequently asked questions

Can ChatGPT write a whole case study from just a client name and website?

No, not one worth publishing. Without a transcript, survey, or real data to work from, ChatGPT will generate plausible-sounding but invented numbers and generic language that could describe any client in any industry, which is the opposite of what makes a case study persuasive.

How long should a case study be?

Most B2B case studies work best between 600 and 900 words on a web page, with a 100-word executive summary at the top for readers who won't scroll past the first screen. Anything over 1,200 words needs a compelling story to justify the length.

Should I use the client's exact words or clean them up?

Stay as close as possible to their actual phrasing, especially in quotes. Over-polished quotes read as fake and clients notice when you've put words in their mouth, which can cost you the relationship even if they sign off on the draft.

What's the biggest mistake people make with AI-written case studies?

Publishing a number the AI generated rather than one the client confirmed. It happens more often than anyone admits, and it's the fastest way to turn a credibility-building asset into a liability if a prospect or journalist ever checks the claim.

Where to check the details

• OpenAI's own platform documentation