A potential customer asks an AI assistant who they should call, and you have no idea what it said. It may have recommended you, described a service you stopped offering two years ago, or sent that buyer straight to a competitor. And the same assistant can name your business today and leave it out of the same conversation next week.
So the useful question is not whether your name appeared once. It is what potential customers are being told about you, whether that information is accurate, and whether the answer sends them your way or to someone else.
If you check one customer question once, you have an observation. If you check a fixed set of questions across the AI tools your buyers use and compare the answers over time, you start to see a pattern. This guide shows what to record and how to connect the findings to the leads and sales you already track.
A quick first check
- In ChatGPT, start a temporary chat and select Unpersonalized before you type, so your saved memories and custom instructions do not shape the answer. In Google, read the AI Overview or AI Mode answer, and click “Try without” in the footer of the results to see a version without personalization. Write down your location and device, because Google says those can still change what you see.
- Ask two questions a new buyer would ask before hiring you, without using your company name, such as “Who helps businesses solve [specific problem] in [your city or market]?”
- Write down whether your business is named, whether the description is accurate, and which competitors are recommended instead.
What does it mean to check what AI tells customers about your business?
Run the same buyer questions on a fixed schedule and record whether your business is named, linked as a source, recommended, described accurately, and shown alongside competitors. Repeat the checks under the same conditions so you can tell a result that keeps showing up from a one-off answer.
The important word is repeat. OpenAI says ChatGPT search results and citations can be incomplete, outdated, or incorrect. Google notes that AI Overviews and AI Mode can use different models and techniques, so the responses and links they show will vary. One answer cannot prove that your visibility has improved or declined.
How do you set up a repeatable check?
1. Start with questions buyers actually ask
Do not begin with your company name. A branded question only tells you whether an AI tool knows who you are. It does not tell you whether it connects your business to the problem a new customer needs to solve.
Build a small list of customer questions around the decisions your buyers make. Include questions such as:
- What are the best [product or service] options for [use case]?
- Who helps businesses solve [specific problem] in [market or location]?
- What should I compare before choosing a [product or service]?
- What is the difference between [category option A] and [category option B]?
- Which businesses are known for [specific capability]?
A family law firm, for example, might ask “Who handles contested custody cases in [county]?” and “What should I ask a divorce lawyer before hiring one?” A dental office might ask “Which dentists near [town] offer same-day crowns?”
Add a few branded questions to check whether your business description is accurate. Keep branded, category, problem, and comparison questions in separate groups. Otherwise, a rise in branded mentions can hide the fact that new buyers still do not come across you.
2. Choose the AI tools that affect your buyers
Start with the tools your customers actually use, not every tool available. Look for evidence you already have: tools customers mention on sales calls, in support questions or in surveys, and AI tools showing up as sources of visits in Google Analytics. If you do not have that evidence yet, pick two tools, such as ChatGPT and Google’s AI answers, and note that the choice is a starting assumption you will revisit.
Google’s own reporting covers part of this. Search Console now has a generative AI performance report that shows how often links to your pages were shown in AI Overviews and AI Mode, by page, country, device and date. It does not show the exact question a buyer asked, the full answer, how your business was described, or which competitor was recommended instead. Testing the questions yourself fills that gap.
3. Keep the test conditions consistent
Record the details that can change the answer:
- Date and time of the test
- Platform, account type, and location, when relevant
- Exact question text and whether web search was enabled
- The full answer or a saved screenshot
- The sources and links shown with the answer
Use the same question wording during each scheduled check. If you change the wording, log the new version as a separate question, or you will not know whether the result changed because of your website or the wording.
4. Record the answer, not just a score
A yes-or-no mention field is useful, but it is not enough. For each response, record:
| Field | What to capture |
|---|---|
| Mention | Was your business, product, or service named? |
| Recommendation | Was it presented as a suitable option, or merely mentioned? |
| Position | Was it first, later in the answer, or buried in a long list? |
| Citation | Did the platform link to your website or a specific page? |
| Framing | What reason did the answer give for including you? |
| Accuracy | Were your location, services, products, pricing, and other important facts correct? |
| Competitors | Which alternatives were named, and how were they described? |
| Action | What source, page, or business fact needs review? |
Save the answer itself when the result is important, because the wording around a mention can reveal a problem a score misses. Being named with an outdated service description is not the same outcome as being named accurately and recommended for the right reason.
5. Repeat the check on a fixed schedule
Monthly testing is a reasonable starting point. A business in a fast-moving category, or one that has recently changed important pages, may want weekly checks for a set period.
Decide in advance how many times you will ask each question during a check, for example three times each, and write that number in your report next to the results. Treat each answer as a sample. Call something a pattern only when it repeats across scheduled checks or across related questions in the same group.
The schedule matters less than the consistency. Run the same list of customer questions, keep the results, and compare the current period with the previous one. That is how to track brand mentions in AI without mistaking one changing answer for a trend. Retire a question only when it no longer represents a real customer question, and record why it was removed.
Get a checklist of what to review on your website so AI tools can find, describe and recommend your business accurately.
Get the AI Visibility ChecklistWhich signals should a business track?
Use a small set of measures that answer business questions.
| Measure | What it tells you | What it does not tell you |
|---|---|---|
| Mention rate | How often your business appears in the answers you tested | Total market awareness or total AI demand |
| Citation rate | How often your website or page is shown as a source | Whether the citation caused a visit or sale |
| Recommendation rate | How often the answer presents you as a suitable choice | Whether the buyer ultimately chose you |
| Accuracy rate | How often the information attached to your business is correct | Whether the platform will update the answer after a correction |
| Competitor presence | Which alternatives appear when you do not | Which competitors appear outside your list of customer questions |
| Visits from AI tools | Visits that arrive with a detectable AI source | Influence from answers that generate no click |
| Leads and sales from those visits | Inquiries, sales, or other actions from those visits | The full value of an answer that influenced a later direct visit |
Keep mentions and citations separate. A platform can name your business based on another website while linking to that other source. It can also cite one of your pages while recommending another business. Those are different problems with different fixes, and why some businesses get named by ChatGPT and others don’t explains where those other sources come from.
Why an AI recommendation and a website visit are not the same thing
An AI answer can influence a buyer without producing a visit you can see in your analytics. The buyer may remember your name, search for it later, walk into your office, buy on another marketplace, or choose not to click at all.
The traffic you can observe still matters. Google recommends using Search Console as the source of truth for how you perform in Google Search and Google Analytics as the source of truth for what people do on your website. Search Console shows how often your pages appeared in Google, how often people clicked them, and which pages and searches produced those visits. Google Analytics can show what visitors did after they arrived. Our guide to tracking visits from AI tools covers how to find that traffic in your own reports.
Do not treat missing source data as proof that AI had no influence. Google explains that a visit can be labeled direct when information about where it came from is missing. Untagged links, redirects, link shorteners, and ad blockers can all contribute to that label.
For the part of the journey you can see:
- Use your monthly list of customer questions to measure mentions, recommendations, citations, accuracy, and competitor presence.
- Use Search Console, including the generative AI report, for how you show up in Google Search.
- Use Google Analytics for visits, engagement, leads, and sales after visitors reach your site.
- Add tracking labels to links you control so Google Analytics can tell where the visit came from, note the limits where you cannot, and compare the measures as related signals.
What should you do when the results change?
The response depends on the type of change, not only the size of the score.
Your business is mentioned, but the details are wrong
Treat accuracy as the first problem. Check your own page and any other website the answer appears to rely on. Correct the underlying business facts, including service descriptions, locations, product details, pricing, and outdated claims. Then repeat the same question after enough time has passed for the platform to pick up the change.
Your business is cited, but a competitor is recommended
Separate the citation result from the recommendation result. Your page may answer part of the question while another business has clearer proof or more relevant supporting information. Review how both businesses are described before deciding whether the gap is your own website, the information about your business on other websites, or a question that does not fit your offer.
Your business is absent from both the answer and the citations
Start with the basics. Confirm that Google can find and list your important pages, and that your own pages link to them. Then review whether your website clearly explains what you do, who you help, and which buyer questions you answer. Google’s guidance is direct: the same basic search practices and people-first content that support Search also apply to its AI features, and there is no special AI markup that guarantees inclusion. For more on that gap, see the difference between a Google ranking and an AI citation.
The score moves up or down for one question
Do not call it a trend yet. Re-run the question the number of times your test method says, compare it with the rest of the question group, and keep the result labeled as an observation until the pattern repeats. A single answer is evidence of what happened in that test, not evidence of a market-wide change.
How should this fit into your business report?
Put the business question first. A report should not open with a mention-rate chart unless the reader already knows why that number matters.
A useful monthly summary might say:
We asked 6 category questions 3 times each this month, for 18 answers. The business was mentioned in 7, cited in 4, and accurately recommended in 3. Visits from AI tools produced 2 qualified inquiries. The next action is to correct the outdated service description appearing in three answers and retest the same questions next month.
Keep the question-by-question log behind it for verification. If you have not run a first check yet, our AI search visibility guide can help structure it.
Key takeaways
- Track brand mentions in AI with the same buyer questions, tools, and test conditions each time.
- Measure mentions, recommendations, citations, accuracy, competitor presence, and visits from AI tools separately.
- Treat one generated answer as an observation, not proof of a trend.
- Fix inaccurate business information before trying to increase the number of mentions.
- Use Search Console for Google Search and Google Analytics for what visitors do on your site, while recognizing that some AI influence will never show up as a visit.
The bottom line
Tracking brand mentions in AI is useful when it shows whether buyers come across your business for the questions they actually ask. It is not useful when it reduces one changing answer to a permanent rank or treats every citation as a lead. The practical way to track brand mentions in AI is to keep the question, answer, source, and date together.
Build a fixed list of customer questions, save the answers, compare results over time, and connect what you can see to qualified inquiries and sales. That shows where AI is helping, where it is sending buyers elsewhere, and which problem to fix next.
Questions to answer about your own business
- When a new customer describes their problem to an AI tool without your name, does your business come up?
- Is what AI tools say about your services, locations and prices still true today?
- Which competitor gets recommended when you do not, and what do those answers say about them?
- Do you know how many of last month’s inquiries started with an AI tool, and how would you find out?
Sources
- OpenAI: Searching the web with ChatGPT
- OpenAI: Memory in ChatGPT (temporary chat)
- Google Search Help: Why your Google Search results differ from others
- Google Search Console Help: Generative AI performance report
- Google Search Central: AI features and your website
- Google Search Console: Performance report overview
- Google Analytics: Understand direct / none traffic
- Google Search Central: Using Search Console and Google Analytics data for SEO
If you want to know what AI tools tell your potential customers and where they send them instead, an Opportunity Review is a call where we look at the evidence available on your business with you.
See What’s Holding Your Marketing Back