What to measure when AI answers mention your brand
A screenshot of your brand in an AI answer can make a strong impression in a budget meeting. Before you fund more content, see how often your business appears when the same buyer questions get an answer. Track source links, identifiable visits, and qualified inquiries as their own measures.
Francisco Contreras · Founder, Machina
6 min read

What can a mention tell you?
Pull up the full response and read the sentence around your company’s name. Did the answer suggest hiring you, mention you in passing, or link to a guide you wrote? If the link goes to an older guide, open it and check whether the information still holds.
We measure where the brand appears by running a fixed set of buyer questions and saving the answers and the pages they cite. Visits from identifiable AI referrers and qualified inquiries tell you what reached the site and the sales team over the same period. Start with a handful of questions your team can run again next month.
Which questions belong in the sample?
Open the last few sales inquiries and look for the questions people asked before they wanted a price. Ask your sales or support team which questions they answer every week. We agree on a set of questions with the client. For each one, write down the decision a buyer is trying to make and the page that should help.
For a hypothetical roofing firm, “Who repairs slate roofs in my town?” tests whether the company comes up when the buyer is looking for options. “Does [company name] repair slate roofs?” tests how the answer describes a firm the buyer already knows. Give each group its own mention rate so the searches that supply your name don’t inflate the number for buyers who are still looking.
Save the exact wording. If a question changes when you launch a new service, add it to the sheet and start a fresh comparison for it. Keep the old row so anyone reading the report can see when the sample changed.
How often does your brand appear when there’s an answer?
Choose where your buyers are likely to ask. Check Google AI Overviews and AI Mode separately alongside ChatGPT, Perplexity, and Gemini wherever they matter to your audience. Google’s documentation explains that AI Overviews can be absent from an otherwise normal results page and that the two Google features can show different answers and links.
Run the same saved questions during the same week each month. Save a copy of each response with the date and any location or account settings you used. Mark a mention when the answer itself names the business and note how it describes you. Copy the linked page, even if it belongs to another publisher. Our guide to AI sources explains how a page can become a citation.
If the answer appears and your name is missing, mark the brand absent. If the results load without an AI answer, write “No AI answer shown.” If the check fails or never gets run, mark it incomplete. Retry failed checks during the same week and keep unresolved attempts out of the completed count.
| Question | Surface | Date | AI answer shown | Brand named | How described | Linked page |
|---|---|---|---|---|---|---|
| Exact wording | Product or feature | Check date | Yes / no / incomplete | Yes or no, only if answered | Recommendation, description, or concern | Visible URLs |
Divide the number of answers naming your brand by the number of AI answers shown. Suppose you plan 22 checks without the brand in the questions. Two fail, eight show no AI answer, and 12 show one. If three of those answers name you, the mention rate is 3 of 12, or 25%. You also know that 12 of 20 completed checks showed an AI answer. Show both counts. If a run turns up zero answers, leave the mention rate blank.
Compare the same questions on the same surfaces from one run to the next. If you miss a set of checks this month, use the checks you have in both months for the comparison. Review the saved answers before deciding which page needs work.
| Month | Question group | Surface | Completed / planned | AI answers shown | Brand mentions | Cited sources |
|---|---|---|---|---|---|---|
| Reporting month | With or without the brand name | Product or feature | Completed checks / planned checks | Answers shown / completed checks | Answers naming brand / answers shown | Answer count by linked page |
Which visits and inquiries can you trace?
Open your analytics traffic-source report and look for referrals from the answer services’ own domains. Group those visits by the page where people arrived. Treat the identified count as a minimum. Visits that arrive without a clear referrer can appear as direct traffic.
Ask sales what makes an inquiry qualified before you pull the total. A roofing firm might count a request for a service it offers at an address it covers. Check that each inquiry reached sales, and remove tests and duplicates. If the records connect a qualified inquiry to a visit from an answer service, include it in the traced inquiry count. Keep “I found you through ChatGPT” as something the buyer told you.
| Month | Referral visits | Qualified inquiries | Traced inquiries |
|---|---|---|---|
| Reporting month | Visits from identified AI referrers | Qualified inquiries from all sources | Qualified inquiries linked to an AI referral |
In Google Search Console, the Generative AI performance report covers AI Overviews and AI Mode together. It measures impressions, which count displays of links to your site. Use the Pages tab and the same reporting dates to see which pages were shown. If the site has too few impressions for the report to appear, keep using your saved answers and referral data.
What did the cases measure?
At Salinas Valley Health, we answered specific patient questions with input from the health system’s physicians and made their clinical credentials visible. Google’s AI Overviews cited one of the articles. Monthly organic visits rose from 22,400 to 163,800 over nine months across the whole site. About 18% of that gain came from AI-related traffic. The number of visits from that one citation remains unknown. To see where an appearance like that can lead, we look at visits to the article and check the path a reader takes to make an inquiry.
In the Union Wine Company case, the team rebuilt category explanations and guides during a site migration. Our founder led this program in a previous agency role, before Machina. Brand mentions in AI answers rose 509% from 2025 to 2026. The case also reported a 125% rise in monthly revenue attributed to AI referrals. The figures leave open whether the increase in mentions contributed to revenue.
Ask what the report counted and over what period. A cited article gives you a page and a patient question to inspect. Referral revenue gives you a reason to look at the pages people arrived on and the products they bought.
Which page should you work on next?
Start where the buyer has a real decision to make. If several answers give the wrong service area, check the address and service details on the pages they cite. If visits arrive at a guide and the path to a quote is hard to find, try that path yourself. Fix the issue you can see, save the date, and run the questions again. Our guide to improving pages for AI search explains how to check access and give the buyer a useful answer.
In our AI visibility work, we agree on the buyer questions, save dated answers and cited pages, and inspect the pages that should answer those questions. We check whether search systems can reach the page and whether the information a buyer needs is there.
If your team needs help running the checks and making the page changes they point to, reach out to Machina. Bring the questions your sales team keeps hearing and the pages you send people to today.
Next steps
Let's talk about your project
Tell us what you're working on. We'll tell you if we can help — and exactly what we'd do.
