The baseline, platform by platform
What an AI visibility audit checks, and what a finding looks like
An AI visibility audit checks six things, each recorded per AI platform, per question and per date: how often and where your business is mentioned, your share of answers against the competitors the AI names instead, the sources cited and the pages cited instead of yours, the accuracy and tone of what the AI says about you, whether the engines can resolve your business as one clear entity, and whether their crawlers can reach the site and lift an answer from it. AI Search Engine Optimization runs the same six checks on its own site every week, across 26 buyer and research questions on six AI platforms, and each card below carries a finding from that record, so you can see the shape of the evidence before you commission it for your own site.
Mentions, and where in the answer
How often each platform names your business when a buyer asks, and the position it gives you: first call, one of five, or an afterthought. Mentions, citations and recommendations are counted as three different things, because they move separately and an agency that reports one number is hiding the other two.
What a finding looks like: across 26 questions on six AI platforms in the seven days to September 16, 2026, AI Search Engine Optimization was named in two answers and cited as a source in twelve, on three pages. Two figures, not one, because a page can be quoted without the business being named, and the gap between them is itself a finding.
Share of answers against the competitors AI names
Of the answers to each buyer question, how many name you and how many name each competitor, with the competitor set taken from the answers themselves rather than from your own list. The businesses the AI recommends instead of you are the field you have to beat, whether or not you thought of them as competitors.
What a finding looks like: in the same seven-day window the three most-named competing agencies appeared in 53, 42 and 35 answers against two for this site, and the field of agencies the AI named ran to eighteen, five of which were not on the original competitor list. The gap is measured to the answer, so the work that closes it can be measured the same way.
Citations, and the pages cited instead of yours
Which of your pages the AI engines quote as a source, and which other pages they quote for the questions you should own: lists, directories, community threads, videos, competitor pages. That second list is the placement plan, because a business earns citations by being on the sources the engines already trust.
What a finding looks like: across the 26 questions on September 16, 2026, Google’s own documentation appeared in 21.7 percent of answers and YouTube in 16.7 percent, ahead of every SEO publication, and a review directory in 9.7 percent. Several answers returned no source list at all; the baseline records those as unmeasured, never as zero.
Accuracy and tone of what the AI says about you
Whether the AI describes your services, your location and your audience correctly, and whether it speaks about you in a neutral, positive or negative register. An inaccurate description is a finding to fix on the sources the AI drew it from, and it is invisible until someone reads the answers instead of counting them.
What a finding looks like: the same buyer question answered on successive runs was localized by ChatGPT to Charlotte, Philadelphia and Santa Fe, with different local agencies added each time, and on September 16 the same question that had named this firm first three days earlier named three larger firms instead. An audit records the location and the date of every answer, or the accuracy check is meaningless.
Entity clarity and structured data
Whether Google, ChatGPT, Perplexity, Gemini and Copilot can resolve your business as one clear thing: a consistent name, address and service list across the site and every profile, one organization record sitewide, and structured data that says what the visible page says. The AI engines name businesses they can verify and skip the ones they cannot, however well the pages rank.
What a finding looks like: when a business name is also its category term, mention counts are unreliable, so the baseline for this site tracks the domain plus two named identifiers, the mechanism name and the founder, instead of the brand words. The audit decides how your business can be detected before it counts anything.
Crawler access and the block an AI can lift
Whether each platform’s crawler can actually fetch the site, probed live rather than read from the robots file, since a firewall can block a crawler the robots rules allow; whether a machine-readable guide to the site exists; and whether each money page carries a passage an answer engine can quote whole, with its schema twin.
What a finding looks like: on September 13, 2026 AI Search Engine Optimization scored 101 answer blocks across eleven of its own pages: none was ready to be lifted whole and 73 were not extractable at all. The rebuild went in the same day. Three days later the first citation of a rebuilt page appeared, taking this site from eight citations on two pages to twelve on three. A number that unflattering is what an audit is for, and what it is measured against afterwards.