The AO Audit · From the scan logs · July 20, 2026
The short version: AI SEO is optimizing your website and profiles so AI engines can find, verify, and recommend your business. It isn't a bag of tricks — each engine answers from an index it can read, so the work is making your business readable: crawler access, schema markup, published prices, consistent profiles. We call the full practice Answer Optimization (AO).
Most articles about “AI SEO” are rewrites of each other. This one comes from a different place: we run a free scanner that checks real business websites for the machine-readability signals AI engines use, and we've already published our own failing grade from it. What follows is what the scans keep showing us — where the engines look, what most sites get wrong, and the order to fix it in.
| Engine | Answers from | What that means for you |
|---|---|---|
| ChatGPT | Web answers lean on Bing's index — citation research traces ~87% of its search citations to Bing | If Bing can't see you, ChatGPT mostly can't either. Claim Bing Places; allow OAI-SearchBot / GPTBot |
| Gemini / Google AI Mode | Google's index + your Google Business Profile | Classic Google hygiene still pays; your GBP is now an AI data source, not just a map pin |
| Perplexity | Its own crawler (PerplexityBot) + cited sources | It quotes pages that state facts plainly — publish answers, not brochures |
Sources for the Bing figure and citation behavior: the SE Ranking citation studies and the 10,000-query GEO research we cite across this site (see footer). The rest of the table is engine-documented crawler behavior.
These come from scanning real local-business websites, not from theory:
Industry-wide, 9 in 10 websites still have no AI instructions file — no llms.txt, no explicit crawler policy. Our scans match that: the single most common result is a site that has simply never told AI engines anything. The field is wide open, which is exactly why moving early works.
The most expensive failure we find. Some site builders render everything with JavaScript in a way that leaves a crawler a nearly empty page — we've scanned a business whose homepage served bots a few hundred characters of nothing. The owner had no idea; the site looked great in a browser. Machine-readable sites earn 3.2× more AI citations in the industry studies we track, and this failure is why: an engine can't cite what it can't read.
Structured data is how you state your name, services, prices, and hours in the format engines parse directly. Most local sites have none. Some have it and it's broken — including, at one point, ours. Valid schema is the cheapest authority signal available.
Customers ask AI full questions: “how much does a fade cost,” “do they take walk-ins.” An engine can only quote what's published. Sites that publish real prices and a real FAQ give the AI something to say about them; sites that hide everything give it nothing — so it names a competitor who didn't hide.
Engines cross-check your website against Google, Bing, Apple, and Yelp. Different hours or an old phone number across those sources reads as unreliable — and unreliable doesn't get recommended.
From the gaps above, in leverage order: (1) unblock and invite the crawlers (robots.txt + llms.txt), (2) install valid schema markup, (3) claim Bing Places and complete your Google Business Profile, (4) publish prices and a genuine FAQ, (5) make every profile agree. That's the spine of every report our audit produces — your specific gaps and order come from your own scan.
What nobody can honestly promise: you can't buy a mention, and no one controls what an AI says on a given day — answers vary by user, session, and engine. Anyone guaranteeing “#1 in ChatGPT” is selling weather. What you can control is whether every engine can read, verify, and quote you — that's the part we measure, and our score counts only what we can verify.
The free AO Audit runs these checks on your site live and hands you an AO Score with your fastest fixes ranked. No account, no card.