We Ran Our Own Playbook: One Day of GEO Work on a Fintech Brand
We applied the full AI-visibility checklist to Pancake, a backtesting platform built by our team. Here is everything that shipped in one day, and what each piece is for.
Disclosure: Pancake is built by the same team as SearchSeal. That is exactly why it makes a good case study — we can show every change, not just the before/after chart.
Most AI-visibility advice is abstract. This post is the opposite: a complete list of what one focused day of Generative Engine Optimization looked like on a real product — Pancake, a platform where AI agents build trading strategies and the platform runs, versions, and hosts them with reproducible results.
The starting point
Pancake already had the foundations many brands skip: an llms.txt manifest, segmented sitemaps, AI crawlers explicitly allowed in robots.txt, and a set of hidden answer pages. What it lacked was coverage of the two question families its future customers actually ask AI assistants: "how do I host a trading strategy?" and "what is the best trading strategy?"
That gap matters more than any technical tweak. AI engines cite pages that answer the question as asked. If nobody on the internet answers your customer's exact question honestly, the engine synthesizes from whoever came closest — and that will not be you.
What shipped, in order of impact
1. Ten question pages for the two unowned intents
Each one leads with a direct answer in the first paragraph, carries QAPage structured data, and is deliberately honest — the "what is the best trading strategy" page opens by saying no universal best exists, then explains what a credible claim would require. Honesty is not a handicap in AI search; it is the citable position.
2. Three entity-definition pages
When you want AI engines to associate your brand with a category, define the category. Pancake now owns a clean definition of "strategy hosting" — the term its product creates. Engines that need a definition have exactly one well-structured place to find it.
3. Three honest comparison pages
Comparative content is the most-cited format in AI answers. Each new page includes a genuine "when to use the other tool" section. We wrote a separate playbook on why that works.
4. A comparative listicle
"How to Host a Trading Strategy in 2026: Six Options Compared" — the format AI engines quote most, written so that five of the six options are competitors described fairly.
5. Machine surfaces for the core artifact
Every Pancake result now serves a Markdown twin at the same URL plus .md, advertises it with a Cite-As HTTP header, and carries a Dataset and Claim structured-data graph. When an agent wants to cite a result, the machine-readable version is one request away.
6. IndexNow ping after deploy
Bing's index feeds ChatGPT Search; the deploy ended with 79 URLs submitted and accepted.
The honest caveat
None of this guarantees citations. It removes every reason an engine would have to cite someone else for questions this product is the right answer to. Measurement comes next — daily citation tracking on the ten target queries, which is precisely the job SearchSeal exists to do.
