AI visibility is the count of times AI assistants such as ChatGPT, Microsoft Copilot and Google’s AI answers name your business or link to your pages when someone asks a relevant question. It is checked in three places: Google Search Console, Bing Webmaster Tools, and a written log of ChatGPT answers plus referral visits. This guide gives one repeatable workflow across all three and shows how to report the result to management.
Key takeaways
- AI visibility measures mentions and citations in AI answers, not positions in a list of search results.
- Google’s generative AI performance report in Search Console covers AI answers in Google Search and has been open to every Search Console account worldwide since the end of August 2026.
- Bing Webmaster Tools’ AI Performance preview reports citations in Copilot and Bing AI answers, and states plainly that those counts are not a ranking.
- ChatGPT has no site-owner report, so it is measured with a fixed question list plus visits tagged utm_source=chatgpt.com in Google Analytics.
- Report the result every month as counts, page lists and enquiries, never as a single AI visibility score.
On this page
What is AI visibility?#
AI visibility is how often AI assistants name your business, describe it correctly or link to one of your pages when a customer asks a question you could answer. There are two things to count: a mention, where your name appears in the written answer, and a citation, where the answer links to a page on your site. Google’s AI-written answers above the search results, called AI Overviews, are one example, and Copilot’s answers in Bing and ChatGPT’s replies are others.
Imagine Eoin, who runs a five-person IT support company in Ballina, County Mayo. His customers now ask ChatGPT or Copilot ‘who does IT support for a small office in Mayo’ instead of scrolling through ten blue links. Eoin’s question is simple: does his firm appear in those answers, and can he prove the answer each month?
AI visibility is not the same as rankings, clicks or the usual search engine optimisation (SEO) reports. A ranking is a position in a list, while an AI answer is written on the spot and may draw on several pages at once. A reader can note a business name from a full answer and never click anything, so click counts and average position miss most of what happened. That is why the checks below combine impressions, citation counts, a written prompt log and referral traffic, and the work of improving them is called generative engine optimisation (GEO).
The monthly loop has six steps and takes about an hour once the question list exists:
- Write down 12 questions customers really ask. Six should include your business name (branded) and six should not (non-branded queries (Searches that do not include your business name)).
- Ask each question in Google, in Copilot or Bing, and in ChatGPT, in a fresh chat each time.
- Record three things per answer: your business named yes or no, your page linked yes or no, and which other businesses were named.
- Read the Google and Bing reports described below for the same month and note their counts.
- Count visits from ChatGPT in Google Analytics and add them to the same page.
- Keep everything in one file so next month can be compared with this one.
Tip: Check the same question list on the same day each month. AI answers shift with wording and with time, so a repeatable check beats a clever one.
Improving the numbers is a separate job, and the guide to ChatGPT SEO covers that optimisation work without repeating it here.
How to check AI search visibility in Google#
Want this done for your business? Talk to us about GEO / AI search optimisation
Google is the only one of the three with a report built for this job, so Eoin starts there. Google introduced generative AI performance reports in Search Console in June 2026, with a separate view of visibility in AI Overviews and AI Mode, two AI features in Google Search, and by the end of August 2026 they were open to every Search Console account worldwide (Google Search Central).
The report shows impressions (The number of times a link or answer was shown) for AI Overviews and AI Mode, and lets a site owner break them down by page, country, device and date (Google Search Console Help). Beside it, the standard Performance report still lists clicks, impressions, average CTR (Click-through rate: clicks divided by impressions, shown as a percentage) and average position for ordinary results (Google Search Console Help).
Read the two together. Set the country filter to Ireland, pick the month, then list the pages that collect AI impressions and compare them with the clicks on the same pages. If a page appears often in AI answers and earns almost no clicks, the answer is doing the talking, so the written prompt log is the only way to learn whether the business was named.
Watch out: A jump in AI impressions is not extra traffic. Pair every impression figure with clicks and enquiries before calling it a win.
Eoin filters to Ireland and finds that his ‘IT support for small offices’ page collects AI impressions while clicks on it stay flat. He adds that phrase to his question list and checks the answers himself to see whether his firm is named.
Does ChatGPT know your business?
Free check in under a minute.
How to use the Bing AI performance report#
Bing covers the Copilot side of the same question, and Eoin opens it straight after Search Console. On 10 February 2026, Bing Webmaster Tools added AI Performance as a public preview, reporting total citations, average cited pages per day, a sample of grounding queries (The searches Copilot used when it built an answer) and page-level citation activity across Microsoft Copilot, AI-generated summaries in Bing and selected partner integrations (Bing Webmaster Blog). The same announcement said citation counts do not reflect ranking or authority, so treat the dashboard as a count of appearances rather than a position table.
In practice, open AI Performance, note the total citations for the month, then read the page list to see which pages are being used. The grounding query sample is the most useful part for an Irish business, because it shows the actual wording people used. Copy that phrasing into the question list and into briefs for whoever writes the pages.
Note: Grounding queries are a sample, not a complete list, so they show how people phrase questions rather than the full size of demand.
Eoin sees that his Microsoft 365 support page is his most cited page, and the sample includes ‘IT support for small offices Mayo’. Those words go into both his page brief and his monthly question list.
How to check if ChatGPT mentions a business#
ChatGPT has no site-owner report in the sources used here, so Eoin measures it by hand and through referral traffic. The manual half is the question list from the first section, asked in a new chat each time and logged with the date, the answer, the mention and any link.
The referral half comes from tagging. OpenAI states that any public website can appear in ChatGPT search if it does not block OAI-SearchBot (The crawler ChatGPT uses to find pages for its answers), and that ChatGPT automatically adds utm_source=chatgpt.com to referral links so publishers can track that traffic in analytics such as Google Analytics (OpenAI Help Center). In Google Analytics, sessions from those links show chatgpt.com as their source, which separates them from ordinary search and social traffic.
Two checks keep the log honest. Ask the questions exactly as written, with no praise or business name added, because a leading question produces a leading answer. Then write the date and the answer next to each question, so an October mention can be compared with a November one without guesswork.
Eoin runs the six non-branded questions first, because a new customer usually starts there. He logs whether his firm is named, whether a link appears, and what the chatgpt.com sessions did on his site afterwards.
How to compare AI visibility across ChatGPT, Copilot and Google#
The three systems report different things, so a fair comparison puts the same month side by side rather than adding the numbers together.
| System | Where the numbers come from | What you get | Main limit |
|---|---|---|---|
| Generative AI performance report in Search Console | Impressions in AI Overviews and AI Mode by page, country, device and date | Impressions only, so pair it with your prompt log | |
| Copilot and Bing | AI Performance in Bing Webmaster Tools | Total citations, cited pages and a sample of grounding queries | Citation counts are not a ranking or an authority score |
| ChatGPT | Written question log plus Google Analytics | Whether the business is named or linked for a fixed set of questions, plus tagged visits | No site-owner dashboard, so the log is a sample |
Compare like with like: same month, same question list, same person doing the check. Count branded and non-branded questions separately, because a name mention in a branded question tells you little, while a mention in a non-branded question is new business. Express the result as counts, for example ‘named in four of twelve questions and linked twice’, rather than as a score out of ten.
The direction of travel matters more than any single month. A page that earns citations in Copilot this month and mentions in ChatGPT next month is doing something right, and the page list shows where to spend editing time.
Eoin keeps one row per month with a column for each system. He uses it to decide which page to rewrite, not to compare Google with Copilot as if they were the same measurement.
How to report AI visibility to management#
The numbers only change decisions when they reach the person who signs off the work, so the report should fit on one page. Eoin brings five lines to his monthly management meeting: total citations from Bing, AI impressions from Google, the share of the non-branded question list where the firm was named, the pages that earned citations, and the chatgpt.com sessions with any enquiries they produced.
Add one sentence on what changed and one on what to try next. Include a short line stating that citation counts are not rankings, because a manager who reads a count as a position will ask the wrong question.
Where the business knows what an average customer is worth, multiply the number of AI-referred enquiries by that internal figure to give a value in euro. Label it clearly as an estimate built from your own sales data, not a measurement of reach.
Eoin’s report takes ten minutes to prepare and ends with one request: permission to rewrite the two pages that never appear in any answer. That is the decision the data exists to support.
How Baikou checks a business in AI answers#
Baikou’s team runs the ChatGPT part of this workflow with its free AI visibility check. It first reads the website the way AI assistants do: the main page, robots.txt and the sitemap. It then asks ChatGPT, with web search switched on, the questions a customer would ask about that type of business in that town, and records whether the business is named, which businesses are named instead and three things to fix first. The answers change from day to day, so the team treats each report as one month’s snapshot and sets it beside the Search Console and Bing figures described above.
Putting AI visibility data to work in an Irish SME#
The next step is to connect the counts to the systems that already run the business, so the data survives beyond one person’s notebook. Three connections do most of the work.
- Analytics: keep the chatgpt.com source intact and check it in the same monthly slot, so AI referrals sit beside the rest of the traffic instead of in a separate spreadsheet.
- CRM (Software that holds customer details, calls and sales notes.): add one field for how the enquiry arrived, with ChatGPT, Copilot and Google AI answers as options, so AI-referred leads appear in the pipeline with a name attached.
- Workflow: give one person the job, one calendar reminder and one shared file holding the question log and both dashboards, with the one-page report going to the management meeting.
Off-the-shelf dashboards show citations and impressions, but they cannot tell you what a customer is worth to your business or why a page was never cited. When the question list grows past a few dozen prompts, or the site runs to thousands of pages, outside help with tracking and page work usually earns its place. Baikou helps Irish businesses with AI visibility and generative engine optimisation work, and the Baikou home page is the starting point.
Free Business Growth Check: see where your business can grow online in 3 minutes →
Want this done for your business?
Talk to us about GEO / AI search optimisation: get recommended by ChatGPT, Perplexity, Google AI.
See GEO / AI search optimisationFAQ#
What is an AI visibility tool?
It is a category of software that runs a fixed set of prompts against AI assistants and logs whether a brand is named or linked. It automates the written log described above, and its value depends on how closely its prompts match real customer questions. It does not replace the Google and Bing reports or your own analytics, because those hold the platform’s own counts.
How can a business track brand mentions in ChatGPT?
Keep a written list of questions, ask them in fresh chats on a set date, and record whether the business is named and whether a link appears. Then check Google Analytics for sessions tagged chatgpt.com and note what those visitors did. Two people using the same log format will produce comparable counts.
What counts as LLM visibility?
LLM stands for large language model, the technology behind these assistants. Visibility is counted in three separate ways: the share of your question list where the brand is named, the number of citations to your pages, and the visits and enquiries that arrive from AI referrals. Report them separately, because each one answers a different question.
Why do impressions rise without obvious clicks?
An AI answer can show a page or a business name without the reader clicking through, so impressions climb while sessions stay flat. The Google report counts appearances and the Bing report counts citations, so neither is a traffic figure. Compare both with clicks and enquiries before deciding that anything changed.
