← Back to the AI visibility service
Live GEO experiment
Engineering a page to get cited by AI on a competitive query
I sell generative engine optimization: making a business one of the names AI engines hand back when a customer asks for a recommendation. This is me running that method on myself, in public, on a query I had no head start on, and documenting exactly how it is built and whether it works.
The test in one line. Take a real, useful data page with zero prior authority, apply the same GEO signals I apply for clients, and see whether ChatGPT, Perplexity, and Google AI begin to retrieve, quote, and attribute it when someone asks "best Indiana scratch-off to buy." No guarantees claimed. Results reported honestly, win or lose.
Why a lottery page
Because it is an honest test. "Best Indiana scratch-off to buy right now" is a real question people ask, it changes weekly as prizes are claimed, and the authoritative data sits on the state lottery site in a form nobody has turned into a clean, current, quotable answer. A new page has no domain-authority advantage there. If GEO signals can move a fresh page into the AI answer for this, the method is real and I can show it, not just assert it. The page itself is genuinely useful on its own merits.
The signals applied
Every one of these is a lever I pull for clients, and the published evidence behind them is in my writeup on getting cited by ChatGPT and Perplexity. On this page they are visible and verifiable.
1. Answer-first, in the first 40 to 60 words
The page opens with the literal answer to the query: the specific games with the most unclaimed top-prize money, and the two to avoid, stated plainly before anything else. AI engines extract and quote the direct answer near the top; burying it under preamble is the most common reason a page gets read but not cited.
2. Structured data that names the answer
The page ships Dataset, FAQPage, and WebPage schema in JSON-LD, with a dateModified and an explicit FAQ answer to "what is the best Indiana scratch-off to buy right now." Schema is how a machine confirms it understood the page the same way a human would, and the FAQ answers are pre-written in the exact shape an engine likes to quote.
3. Clear authorship and a retrievable source trail
The data cites its origin (the official Hoosier Lottery prize report) and names its maintainer, linked to my Person entity that the rest of the site already establishes. Engines weight sources they can attribute to a named, consistent entity. When the answer gets quoted, the citation points back here.
4. Freshness as a first-class signal
A "last updated" date on the page, in the schema, and in the sitemap. For a query whose correct answer changes weekly, recency is not cosmetic; it is the difference between a right answer and a wrong one, and engines prefer the source that is visibly current.
5. Machine-readable context files
The page is registered in llms.txt, the plain-text index built specifically for AI crawlers, and mirrored in a screen-reader-only context block that restates the key facts in prose. Both give a retrieval system a clean, unambiguous version of the answer.
What I am measuring
| Check | How |
|---|---|
| Is the page cited? | Ask ChatGPT, Perplexity, and Google AI the buyer questions, screenshot whether this page is named or linked |
| How fast? | First check at two weeks, again at 30 days, since engines re-crawl on their own schedule |
| Which signal mattered? | Compare against the incumbents that rank but are paywalled or app-only, and note what this page has that they do not |
The honest caveats
- AI answers are regenerated per conversation and vary run to run. "Cited" means it shows up reliably across repeated, varied prompts, not once.
- Getting cited is the top of the funnel, not revenue. AI answers are often zero-click. The value here is the demonstration, the domain authority it lends the rest of the site, and the proof I can put in front of the right prospect.
- If it does not work, that goes here too. A method you only hear about when it wins is a sales pitch, not a case study.
Why this matters for a business
When your customer asks an AI engine "who's the best [your category] near me," the engine names two or three businesses. If you are not one of them, the customer never hears you. This page is the same set of moves I make to get a business into that answer, applied where you can inspect every one of them. If that is the outcome you want, that is the AI visibility service, and the free check is the whole pitch.
Want your business named by AI, with the proof up front?
I run your customers' questions through ChatGPT, Perplexity, and Google AI, screenshot who gets named today, and show you the fixes. If AI already recommends you, you owe nothing.