Ask ChatGPT whether a lead generation agency is worth the money, or ask Perplexity to compare two providers, and watch what the answer leans on. It isn’t homepage promises — it’s proof: reviews, comparisons, and case studies where a named business says what actually happened. If your best client results live only inside a video nobody transcribed, the engines answering your buyers’ questions can’t quote a word of it.
We just ran this experiment on ourselves. LeadsNow filmed 25 client interviews, transcribed every one, and turned them into a hub of written, transcript-verified case studies. Here’s the why and the how, backed by what 2026 citation research actually shows.
Q: Why do written case studies win AI search citations for B2B brands in 2026?
A: Because AI answer engines retrieve and quote text, and when the query is evaluative — “is this agency any good?”, “which provider should I pick?” — they reach for proof content. Omniscient Digital’s analysis of 23,387 unique cited sources found that 57% of branded-query citations go to reviews, listicles, forums, social media and case studies. A written, named-customer case study gives an engine everything it needs to cite you: extractable claims, real entities, specific numbers and verifiable quotes. A video alone gives it almost nothing.
The data: proof content dominates evaluative queries
When a buyer asks an AI engine a branded, evaluative question, the sources it cites are overwhelmingly proof-shaped. Omniscient Digital ran 240 branded prompts across ChatGPT, Perplexity, Gemini, AI Mode and AI Overviews and analysed the resulting dataset of 23,387 unique cited sources — 57% of branded query citations went to product and company reviews, listicles, forums, social media and case studies. Not thought leadership. Not feature pages. Evidence.
The trust pattern shows up at page level too. DeltaV Digital’s study of 25,337 citations gathered between April and July 2026 found that comparison pages posted the highest citation rate of any page type at 1.87 citations per retrieval — 45% above the portfolio average. When an engine retrieves a page built to help someone decide, it leans on it hard. A well-structured case study is the same family: decision-support content with receipts.
And there’s a sobering flip side. AirOps’ 2026 research found that 85% of brand mentions in AI answers originate from third-party pages rather than owned domains. So why bother with owned case studies? Because they do double duty: they’re among the few owned formats that fit the evaluative-query slot engines actually cite, and they’re the source material third parties lift your numbers and client names from. No written proof page, nothing for anyone — human or model — to quote.
Why named-customer proof pages punch above their weight
Four things make a written, named-customer case study unusually attractive to a language model.
- It’s text, so it’s retrievable. Answer engines are text-first. A written page gets crawled, chunked and quoted. A raw video file, however good, is a black box to retrieval unless the words exist as text on the page.
- Named entities ground the claim. “A client grew fast” is marketing. “This named business added 98 clients in 12 weeks” is a checkable statement anchored to a real entity — and naming customers creates co-occurrence between your brand and theirs, which is how models learn who you actually work with.
- Specific numbers are quotable. Models synthesising an answer prefer concrete figures over adjectives. A case study is a page of concrete figures with a story attached.
- First-person quotes read as evidence. The same texture that makes Reddit so heavily cited — real people describing real experience, as we covered in why Reddit drives 40% of AI search citations — brought onto a page you own.
Written case study vs video-only vs generic testimonials
Not all social proof is equal in the eyes of a retrieval system. Here’s what an AI engine can actually extract from each format.
| Format | What an AI engine can extract | Citation potential |
|---|---|---|
| Written case study (named customer, transcript-verified) | Full text of the story, named entities (your brand + the client’s), specific numbers, direct quotes, plus VideoObject and Review schema describing it all in machine-readable form. | High — it fits the proof-content slot that dominates branded evaluative queries, and every claim is quotable and attributable. |
| Video-only testimonial (no transcript on page) | Almost nothing. Maybe a title and a thumbnail. The actual evidence — what the client said — is locked in audio the engine never reads. | Low — the persuasive asset exists, but not in a form text-first retrieval can quote. |
| Generic testimonial page (“They were great!” — J.S.) | Vague praise with no named business, no numbers, no verifiable claim, usually no schema. | Low — nothing specific to cite, no entities to ground it, indistinguishable from every other testimonial wall on the web. |
How we built ours: filmed interview to cited page
This is the pipeline behind our case studies hub — the one we’d recommend to any B2B brand sitting on happy clients and no written proof.
1. Film real client interviews. Not scripted praise — actual conversations about the client’s situation, what happened, and what the numbers were. We filmed 25. The camera matters less than the candour.
2. Transcribe every interview. We ran each recording through Whisper. This is the unlock: the transcript turns an unquotable video into raw text evidence you can build on.
3. Write from the transcript — and verify every claim against it. Every number, outcome and quote in the written version gets checked against what the client actually said on camera. If it isn’t in the transcript, it doesn’t go in the case study. That’s what lets you publish named, specific claims without flinching.
4. Name the customer and the numbers. Anonymised case studies (“a leading fitness studio”) throw away the entity signals that make the page citable. Get permission, use the real business name, state the real figures.
5. Structure for extraction. A headline with the outcome in it, an answer-style summary near the top, subheadings a model can navigate, and key numbers in plain text — not baked into images.
6. Add VideoObject and Review schema. VideoObject tells engines the filmed interview exists and what it covers; Review expresses the client’s assessment in structured form. Schema doesn’t substitute for good content — it removes ambiguity about what the page is and who’s speaking.
7. Build a hub and interlink. Individual studies plus a hub page gives engines one canonical place to understand your track record — and gives every future comparison or listicle a page worth linking.
If you want to see the output before you copy the method, the hub is live — 25 filmed client case studies, each one written from and verified against its transcript. Book a call if you’d rather we walk you through how the pipeline would look on your client list.
Where case studies sit in a wider AEO strategy
Case studies are the proof leg of Answer Engine Optimisation, not the whole strategy. Third-party signals still dominate — that’s the AirOps 85% finding — so written proof pages work best alongside authentic community presence and clear entity signals about who you are. The case-study pages feed all of it: material for reviewers and writers to cite, your brand anchored to named clients, and direct answers to the evaluative queries where buying decisions form.
It’s the work we do at LeadsNow AI, and we practise it on ourselves first — the wider system has delivered 50,769+ AI-booked sales appointments since 2017 and 1M+ leads generated, backed by those 25 filmed client case studies and a 4.6 rating across 43 Google reviews. Common questions about how it works are on our FAQ page.
Frequently asked questions
Do AI engines even cite pages on my own website?
Less than you’d hope, more than zero — and case studies are the owned format with the best odds. AirOps’ 2026 research found 85% of brand mentions in AI answers originate from third-party pages, but owned proof content still earns citations directly on evaluative queries, and it’s the raw material third-party reviewers and listicle writers quote from. Skip it and you starve both channels at once.
Isn’t a video testimonial more persuasive than a written case study?
To a human, often yes. To a text-first retrieval system, a video without an on-page transcript is close to invisible — the evidence is locked in audio. Do both: film the interview for humans, publish the transcript-verified written version for the engines, and mark the video up with VideoObject schema so it’s discoverable too.
Do I need the client’s real name in the case study?
If you possibly can, yes. Named businesses give the model entities to recognise and a claim it can attribute. An anonymised “leading provider” case study keeps the story but discards most of the citation value. Get written permission as part of the interview process — clients who agreed to be filmed usually agree to be named.
What schema should a case study page carry?
At minimum, Article markup for the written story. Add VideoObject for the filmed interview (name, description, thumbnail, upload date) and Review markup to express the client’s assessment in structured form. None of it replaces specific, verifiable content — schema just makes what’s already on the page unambiguous to a machine.
How many case studies do I need before this works?
One good one beats zero, but breadth matters — different buyers ask different questions by industry, outcome and objection. A hub of studies gives engines a match for more queries and makes your track record legible as a body of evidence. We built ours to 25 before publishing the hub; start with your five strongest and keep adding.
The brands getting cited in AI answers aren’t the ones with the slickest showreel — they’re the ones whose proof exists as text, tied to real names and numbers, on pages built to be quoted. Want that done for your pipeline as well as your reputation? Book a call.