Guide
What Is AI Visibility?
A plain-English guide for business owners who want to understand why AI engines recommend some businesses and not others — and what to do about it.
The short version
When someone types a question into ChatGPT, Gemini, or Google AI — "best estate planning attorney in Denver" or "affordable HR software for small teams" — the AI engine generates a response recommending one or more businesses. AI visibility measures how often your business is among those recommendations, and how prominently it appears when it is.
A business with high AI visibility is regularly recommended across multiple AI engines for the searches that matter to it. A business with low AI visibility is either absent from those responses or mentioned in passing while competitors are recommended more confidently.
How it differs from SEO
Traditional SEO optimizes for search engine rankings — getting your web pages to appear at the top of Google results for specific keywords. The goal is a high-ranking page. The measure of success is click traffic.
AI visibility optimizes for recommendation engine responses — getting AI engines to mention and endorse your business when someone asks a natural-language question. The goal is being named as an answer. The measure of success is appearing in the response at all, and how positively you're described when you do.
The two are related, but they are not the same measurement: a Google ranking is about where a page appears, and an AI answer is about which businesses get named.
Where an answer actually comes from
It is easy to picture this as an engine reading your website at the moment somebody asks. That is not the usual path — Perplexity, which documents fetching a page when a user asks, is the exception — and knowing what happens instead explains some of what looks strange in your results.
Pages from the index, or a page fetched seconds ago. Fresh. This is where citations come from, and the only part you can change this quarter.
Absorbed when the model was trained, months or years ago, from an enormous amount of text. No citation, because it cannot say where a fact came from.
Separately from all of the above, training crawlers collect web text that is used to build the next model. That is where “what it already remembers” came from — and it is why nothing you publish this week touches it.
Two things follow from this that are worth holding on to. A business can be named often and cited never — that answer is coming from what the model already knows rather than from your site. And unblocking a crawler does not produce an immediate result: the door opening is not the same as the crawler having come back through it.
Can you do anything about what the model already remembers?
Not in any way you can measure. It is worth understanding anyway, because it explains results that otherwise look contradictory.
No AI company documents what puts a business into a model's memory, and we have not measured it. Our understanding is that it comes from being written about in many independent places over time, but that is our view, not something we can show you.
Allowing the training crawler is a choice, not a lever. For ChatGPT, OpenAI documents that its training crawler, GPTBot, is separate from its search crawler, and that blocking GPTBot does not remove a site from ChatGPT search.
And whatever the payoff is, it lands when the next model is trained, which no AI company publishes a date for. We would be wary of any service that promises to get your business into an AI's training data: we do not know of a way to measure whether it worked.
A trade publication write-up, a supplier case study naming you, a local news piece, an association profile — each is a page a search crawler can read and an engine can cite now, and it may also contribute to a future model. So we would judge that work by what you can measure.
What we would not do is spend money chasing the training front on its own. Judge the work by the front you can see.
Which AI engines matter
OpenAI's AI assistant.
Appears directly inside Google search results.
Google's standalone AI assistant.
Anthropic's AI assistant.
An AI search engine that shows its sources.
The weights are our estimate of each engine's share of consumer AI search. Which sources get cited may depend on the kind of business. AirOps and Foundation's 2026 study of software searches found review sites like G2 among the most-cited sources; in our scans, mostly of local home-technology businesses, most citations were ordinary websites, mostly those of the businesses being recommended.
What affects AI visibility
Below is what the engines document, or what we have measured, with the scope of each. Some things commonly recommended are not on the list because we have not measured them: reviews, press mentions, and consistency across sources.
Crawler access to your site
Documented: OpenAI states that sites which disallow its search crawler, OAI-SearchBot, do not appear in ChatGPT search answers, and that a robots.txt change takes about 24 hours to register. A firewall can also block a crawler by name, and we check. It is a failure that gives no symptom you would notice: the site looks fine to you, because your browser is not the visitor being turned away.
Text on a page an engine can read
Documented: ChatGPT search, Perplexity and Google AI Overviews build answers from pages they retrieve, so a fact must exist as text on a page they can read to be quoted. Text in images, or drawn in by scripts a crawler does not run, is not available to them. If customers ask "how much does X cost" and no page of yours answers it, an engine has no page of yours to quote for that question.
Being indexed by Google
Documented: to appear in a Google AI Overview a page must be indexed and eligible to show in Google Search with a snippet. Google says there are no additional technical requirements, and no special schema.org markup is needed.
Websites are most of what gets cited
Measured: across 3,490 citations in our scans to 5 September 2026, mostly home-technology searches in the US and Canada, ordinary websites, mostly those of the businesses being recommended, were about 86% of ChatGPT's citations, 85% of Gemini's and 73% of Google AI's.
Google Business Profile
Measured: in the same scans, Google AI Overview cited Google Business Profile cards in 12.3% of its citations. The card carries the profile's description, rating and review count; a local business without a profile has no such card to be cited. We have not measured the effect of review recency or volume.
Certification and trade-body listings
Measured: on the home-technology searches we scan, certification and trade-body sites such as CEDIA were 4.7% of all citations and 9.9% of ChatGPT's (9 September 2026, 6,047 citations). A complete listing on a body you already belong to is a page that names you on a domain the engines cite.
How to measure it
Measuring it requires asking each AI engine the same questions your customers would ask — and recording whether your business was mentioned, how it was described, whether your website was cited, and how you compared to the competitors who did appear.
Doing this manually across five AI engines, multiple search phrases, and multiple competitors quickly becomes impractical. Automated tools that run these queries on a regular schedule and track changes over time give you the trend data needed to know whether what you're doing is working.
How long improvements take
| Action | When to check back |
|---|---|
| New Google reviews (especially detailed ones) | 2–4 weeks |
| Google Business Profile updates | 2–4 weeks |
| Adding a pricing or FAQ page | 3–6 weeks |
| Updating service descriptions with customer language | 4–8 weeks |
| Building directory citations and listings | 6–12 weeks |
| Press mentions or industry publication features | 8–16 weeks |
These are our estimates of when it is worth looking again, not promises of when something will move. No AI engine publishes a recrawl timetable, and we have not measured these against real results. The one documented timing is for crawler access: OpenAI says a robots.txt change takes about 24 hours to register. Tracking your score over time is how you see whether a specific change had an effect.
See where your business stands
A free indextr scan checks whether ChatGPT names your business, calculates a Visibility Score, and tells you what to work on, with no credit card required. Starter adds Google AI Overview and Gemini; Pro covers all five engines, adding Claude and Perplexity.
Get Your Free Score →