← All posts

September 25, 2026 · 7 min read

If Your Website Doesn’t Speak "AI," AI Skips You

If an answer engine cannot read a clean, specific claim off your website, it will not guess on your behalf. It moves to the next business whose page states the service, the price range, the city and the hours in plain text — and that business is often a franchise or a national brand with a marketing budget you do not have.

That is what people mean when they say a site does not "speak AI." It is not a tone of voice or a new keyword trick. It is whether the facts about your business exist on your pages as text a machine can lift and attribute. Below is what makes a site unreadable to ChatGPT, Google AI Overviews, Perplexity and Gemini, in the order it usually matters, and what to change.

What “speaking AI” actually means

An answer engine does three things before it names you: it fetches your page, it extracts statements it can treat as facts, and it decides whether those statements are specific enough to repeat with your name attached. Fail any one of the three and you are not in the answer.

None of that requires clever writing. It requires that the important things about your business be written down. A page that says "we deliver world-class solutions tailored to your needs" contains no facts. A page that says "we repair sliding glass door rollers in Palm Coast, Ormond Beach and Daytona Beach, typically $250 to $600, same-week appointments" contains five.

Most local business sites are not badly written. They are vague, and vagueness is what gets skipped. Somewhere along the way the industry taught owners to sound like a brand instead of sounding like a business with prices, service areas and hours.

The four ways a site becomes invisible

In audits, the same four problems come up again and again — and they stack. A site with three of them is effectively absent from AI answers no matter how good the work behind it is.

  • Content that only exists after JavaScript runs. If your text is assembled in the browser, a crawler that reads raw HTML may see an empty shell. Server-rendered text is not optional anymore.
  • Facts trapped in images. Prices, service lists and hours set in a graphic or a PDF are unreadable. A machine cannot quote a JPEG.
  • No structured data. Schema markup is how you hand over your business type, location, services, hours and FAQs without asking a model to infer them from layout.
  • One page trying to cover everything. A single "Services" page listing nine services gives an engine nothing specific to cite for any one of them.

Notice that three of the four are structural. This is why "just write more blogs" rarely moves the needle on AI visibility. If the foundation cannot be read, more content on top of it cannot be read either.

Search Console chart showing AI answer impressions rising month after month
A client account: appearances inside AI-generated answers climbing after the pages were rewritten to answer questions directly.

Why the answer usually goes to the bigger company

When a model has no strong local source, it falls back on what it does have: large sites with consistent structure, heavy schema, deep page counts and a lot of external mentions. That is a description of a franchise, a directory or a national chain.

The result feels unfair, and in a sense it is — the local operator often does better work. But the mechanism is not a preference for big brands. It is a preference for legible, verifiable sources. That is a gap you can close, because legibility is cheap and it is not something a national brand can hoard.

The good news

Specificity is the one advantage a local business has that a national brand structurally cannot match. Nobody in Atlanta can write a more credible page about roof repair in Flagler Beach than the crew that works there — as long as the page actually says Flagler Beach.

What a page that speaks AI looks like

Take one service page and rewrite it against this shape. It is not complicated, and it works for both search engines and answer engines at the same time.

  • One H1 naming the service and the place: "Sliding glass door repair in Palm Coast, FL."
  • A first paragraph that answers the obvious question outright — what it costs, how long it takes, whether you serve the caller’s city.
  • Question-shaped H2s that mirror how people ask: "How much does it cost?", "How soon can you come out?", "Which cities do you cover?"
  • Direct answers in the first two sentences under each H2. Machines lift openings, not conclusions.
  • Numbers wherever you can defend them: price ranges, warranty length, years in business, response window.
  • A short FAQ block with four or five real questions, marked up with FAQ schema.
  • Named service areas in text, not only on a map embed.

Do that once and you will notice something: the page also converts better for humans. Nobody has ever left a service page because it told them the price too early.

What you cannot fake

There is a temptation to treat AEO as a markup game — bolt on schema, add an FAQ block, done. Structured data helps, but it is a label on a box. If the box is empty, the label does not help.

Answer engines cross-check. They compare your page against your Google Business Profile, your reviews, your listings and mentions elsewhere. Claims that contradict those sources tend to get dropped rather than repeated, because a model that quotes wrong information about a local business gets embarrassed in a way it is trained to avoid.

So keep your name, address, phone, hours and service areas identical everywhere they appear, and never publish a price or a guarantee you would not honor on the phone. Consistency is a ranking asset now, not just good hygiene.

How we handle this on every build

We do not sell AEO as a separate line item, because it is not separate work. Server-rendered text, one page per service, city coverage where it makes sense, schema, sitemap, robots, an llms.txt file and answer-first copy are how we build sites — at $300 for the first five pages, $50 per extra page, and $50 a month for hosting.

There is no retainer, no sales call and a 30-day money-back guarantee. The AI part of our own operation is why the pricing works: software drafts and produces, humans in the US decide what is true, what to say and what ships.

If you want to see where your current site stands before you spend anything, run the free audit. It reports SEO, AEO, UX and page speed, and it tells you which of the four invisibility problems above you actually have.

Search Console chart with flat impressions for months then a steep climb
Same site, total search impressions: flat for months, then a steep climb once the structure changed.

FAQ

Questions people ask about this

Does my website really need to change for AI search?
Usually yes, but less than people expect. The changes are structural — server-rendered text, schema, one page per service, answer-first copy. They also help traditional rankings, so it is not work you do twice.
Will AI skip my business if I only have a Google Business Profile?
Often. A profile gives an engine your name, hours and reviews but nothing quotable about how you work, what you charge or what you fix. The website is where the citable detail lives.
Is adding schema markup enough?
No. Schema labels facts that already exist in your content. If the page never states prices, service areas or specifics, the markup has nothing meaningful to describe.
How do I tell whether AI mentions my business?
Ask it. Type your service plus your city into ChatGPT, Perplexity and Google and see who gets named. Repeat it monthly. It is crude, but it is the most honest measurement currently available.
How much does it cost to make a site AI-readable?
With us it is included: $300 for a five-page build, $50 per additional page, $50 a month for hosting. No separate SEO or AEO retainer.

Check the basics yourself

Run a free audit of your current website

Titles, descriptions, page speed, mobile layout, and how readable your pages are to AI assistants — scored in a couple of minutes, with no email required.

Keep reading