How to Write a Case Study with ChatGPT: A Prompt Workflow

A case study is the most persuasive content most businesses never write, because it looks expensive to produce. It is not, if you already have the raw material: a customer, a problem, and an outcome. ChatGPT will not invent those facts — and it should not — but it will structure, draft, and tighten them far faster than a blank page.

This five-step workflow takes scattered facts and turns them into a case study that reads like a real story with a real number at the end.

What You Will End Up With

A publish-ready case study in the classic problem → solution → result shape, with the customer’s own words in the right places and every metric traceable to a source you provided.

Step 1: Gather the Facts First

Never let the model fill gaps. Feed it only what is true, and label what is missing.

Here are the facts I have about a customer project: customer type, the problem they had, what we implemented, the timeline, and the results. List what is present, and separately list every gap I still need to fill before writing. Do not invent anything.

<paste your facts here>

You get an honest inventory. The “gaps” list becomes your interview checklist — this is the step most people skip and later regret.

Step 2: Build the Outline

Lock the structure before drafting, so the draft does not wander.

Create a case study outline with these sections: Snapshot (customer + one-line result), The Challenge, The Approach, The Results, and a pull-quote slot. Under each section, write the one key message it must land.

You get a skeleton you can approve in seconds, before committing to full prose.

Step 3: Draft the Narrative

Now expand, section by section, in the customer’s situation rather than yours.

Write The Challenge and The Approach using only the facts provided. Open with the customer's situation, not our product. Keep it to 250 words, past tense, no hype adjectives. Where I gave a quote, place it verbatim.

You get a draft that reads like a story about the customer, which is exactly what makes it convincing.

Step 4: Add the Results and Quotes

Results need specificity and honesty. Quotes need to be verbatim.

Write The Results using only these numbers: <list metrics>. State each with its source. If a number is missing, write "[metric pending]" instead of guessing. Suggest one place where a customer quote would strengthen the section, and tell me what the quote should be about.

You get a results section that is credible because every figure is sourced, and a targeted ask for the quote you need.

Step 5: Polish and Trim

Cut the filler that creeps in during drafting.

Tighten the whole draft to 700 words. Remove adjectives that add no information, convert passive voice to active where possible, and make sure the first sentence states the outcome.

You get a tight final piece where the payoff appears early — the way busy readers actually consume case studies.

Final Deliverable Checklist

• Every fact traced to a source you provided; no invented numbers.
• Classic structure: Snapshot, Challenge, Approach, Results, quote.
• At least one verbatim customer quote.
• Outcome appears in the first sentence.
• Final length appropriate to your format (often 600–900 words).

Common Pitfalls

Pitfall 1 — letting AI invent metrics. A case study with a fake number is a liability. Always mark unknowns as pending.

Pitfall 2 — writing about the product. Readers care about the customer’s before-and-after. Keep the product in a supporting role.

Pitfall 3 — burying the result. Put the outcome first; the hard numbers are the hook, not the ending.

If you are still shaping the writing, start with our ChatGPT blog post workflow. To cut the filler, see ChatGPT rewrite prompts, and browse more angles in our ChatGPT writing prompts list.

Worked Example: The Skeleton, Filled

A finished outline shows how little prose you need before the draft.

Snapshot: A 12-person logistics startup cut invoice processing from 6 days to 8 hours.
The Challenge: Invoices arrived by email in three formats; two staff re-keyed them by hand.
The Approach: We standardised intake, added an AI extraction step, and kept a human check on exceptions.
The Results: 6 days -> 8 hours; 40% fewer manual hours; error rate from 4% to under 1%.
Pull-quote slot: a sentence from the operations lead about what the time saving meant for the team.

With the skeleton approved, drafting step 3 becomes mechanical — which is exactly the point.

Where to Publish a Case Study

Interview Question Bank

Use these to fill the gaps the outline exposed:

• What was true before you started that you wanted to change?
• What did you try first, and why did it not work?
• What number mattered most to you, and what was it before?
• What would you say to another team considering this?

Quick FAQ

How long should a case study be? 600-900 words for a web page; a one-slide snapshot for a sales deck. Match the format to how the reader will actually meet it.

What if the customer will not share numbers? Lead with the qualitative change and mark every figure as pending. A case study without a fake metric is still persuasive; one with a fake metric is a liability.

Should I name the customer? Only with written permission. Otherwise use a descriptor such as “a 12-person logistics startup” — the story still works without the logo.

Can I turn one case study into more content? Yes. The snapshot becomes a social post, the challenge section becomes a blog intro, and the results become a sales email. One interview can feed a month of content.

Sources & Notes

The problem-solution-result case study structure is a common marketing and B2B convention; the wording and prompts here are our own, reviewed on 2026-10-09. Always obtain written permission from the customer before publishing their name, data, or quotes, and have legal review any performance claims before release.