How AI engines decide
what to say about you.
Ask an AI engine about a business and it composes an answer from everything public: the site's raw bytes, the directories, the socials, the third-party pages. It rewards consistency. This practice has run that play, kept the receipts, and will show them.
Discipline
AEO · GEO
Case
Documented, 2026
Free pass
Homepage, 5/day
Baseline
$500, credited
- The receipts
One case,
traced end to end.
In spring 2026 this practice ran a full visibility engagement for a Louisville catering company: one authored phrase-set, deployed consistently across the website, a scheduled social program with a published record, and directory listings carrying the same sentences.
In September 2026 - months after the engagement closed and the website itself had been replaced - both Google's AI results and Perplexity still described the business in that authored language. One authored sentence appears word for word in a directory listing the AI engines cite. The trace runs from source file to published post to live AI answer, with the publish records to match.
The same audit also found the failure mode: schema markup that only existed after JavaScript ran, which no AI crawler ever saw. Both lessons are built into how this site itself is served - and both are checkable right now with view-source.
- Asked plainly
The AI visibility
questions, answered.
How do AI engines decide what to say about a business?
They compose an answer from everything public about you - your site as raw fetched bytes, directory listings, social profiles, third-party pages - and they reward consistency. When the same clear sentences about a business appear across many surfaces, those sentences tend to become the answer. When surfaces disagree, the engine averages, hedges, or picks a competitor.
Can you guarantee what ChatGPT or Google AI says about my business?
No, and no one can - walk away from anyone who promises otherwise. What can be engineered is the input: true facts, stated consistently, in crawlable bytes, distributed to the surfaces the engines actually read. In a documented case this practice ran, language authored in spring was still composing two AI engines’ description of the business months later.
Does schema markup matter for AI search?
Only when it is in the bytes. Most AI crawlers do not run JavaScript, so schema injected at runtime is invisible to them - this practice proved that on a live site where a rendering validator found nothing because the markup only existed after script ran. Static structured data, served in the page source, is the version that counts.
What is llms.txt?
A plain-text index at yoursite.com/llms.txt written for AI crawlers - who you are, what you do, where the substance lives - in the format they parse best. This site serves one, plus a full-register version, and explicitly allows the major AI crawlers in robots.txt. Most sites still serve neither.
How long before AI answers reflect the work?
Months, not days. The documented case placed content in spring and the AI surfaces were still composing from it in September - after the original website had been replaced. Distribution outlives any single page, which is exactly why the work front-loads consistency across surfaces rather than tweaking one homepage.
What is a GEO audit?
Generative Engine Optimization - how your business surfaces in ChatGPT, Perplexity, and Google AI Overviews, not just the blue links. The structural pass runs free on this site’s homepage, five runs a day. The five-hundred-dollar Baseline adds the citation map across the AI surfaces and the strategic narrative, and credits against a build.
The structural pass runs free on the homepage, five runs a day. The Baseline maps your citations across the AI surfaces and credits against a build.
Request the audit · 502-345-0525 · Who actually builds custom?