A customer tells you they “asked ChatGPT for a good shop in town” and your name didn’t come up. That’s a marketing problem now, and unlike a ranking drop, no dashboard tells you it happened. This guide covers how to track brand mentions in AI search without an enterprise platform: the prompts to test, what to write down, the free reporting Google gives you, and the point where paying for a tool makes sense. If you also want to change what the answers say, start with our guide to ranking in AI Overviews.
Why you need to track brand mentions in AI search
For twenty years the question was “where do we rank?” and one tool answered it. Now a chunk of your category’s research happens inside an answer. Someone asks which shops handle aluminum repair, or whether a panel upgrade needs a permit, and gets back three named businesses and a short explanation.
Two things make this hard to see. The answers are personalized and non-deterministic, so the same prompt can name different businesses an hour later. And most of the engines report nothing back to you. You end up measuring it the way you’d measure word of mouth, with a fixed sample that you check the same way on a schedule.
Mentions, citations and recommendations are three different things
If you lump them together, the report won’t tell you much, so track each one separately:
- A mention is your business name appearing in the answer text. It helps with awareness, but there’s no link and it’s hard to attribute.
- A citation is your page listed as one of the linked sources under or beside the answer, and it’s the one that can send traffic.
- A recommendation is the answer actively putting you forward as a choice, usually in a shortlist. That’s what produces calls.
You can be cited without being recommended (the engine used your pricing guide, then recommended a competitor) and recommended without being cited (it knows your name from directories but linked elsewhere). The gap between the two shows where the work is. A citation gap is a content problem, while a recommendation gap usually comes down to reputation and third-party mentions.
Step 1: write down the ten prompts your customers ask
Write them as prompts rather than keywords: the full, clumsy sentences people type when they’re talking to a machine. Pull them from the questions your front desk answers, your email inquiries and the chat transcripts on your site.
Aim for a fixed set of ten to fifteen, split across the funnel:
- Category shortlists: “who are the best collision repair shops in Tulsa?”
- Problem-first: “my water heater is leaking from the bottom, who do I call in Denver?”
- Comparison: “paintless dent repair vs traditional bodywork, which should I choose?”
- Qualifying: “which recycling companies near me handle commercial roll-off?”
- Branded: “is [your business] any good?” and “[your business] vs [competitor]”
Freeze the wording. The method only works if you ask the same thing every month, because then a change in the answer means something.
Step 2: run the prompts and record the same five things
Run each prompt on the surfaces your customers use: ChatGPT, Google AI Overviews and AI Mode, Perplexity, Gemini, and Copilot if your customers are on Windows desktops all day. Use a logged-out or fresh session where you can, so you’re not measuring your own history. Note the date too, because model versions change underneath you.
Then record five columns and nothing more:
| Column | What you write down |
|---|---|
| Included? | Yes or no: were you named in the answer at all? |
| Position | First, middle or last in the list of businesses named |
| Described as | The exact phrase used about you, such as “budget-friendly” or “specializes in fleet work” |
| Cited? | Which of your URLs, if any, appears in the sources |
| Who else | Every competitor named, in order |
The sheet gives you two numbers without any tooling. Your share of answers is the percentage of prompts where you were named at all, and your citation share is how often one of your pages was among the sources. Both are only useful as a trend, so treat the first run as a baseline. Don’t skip the “described as” column. It’s how you find out the internet thinks you’re the cheap option.
Step 3: check which sources the answers lean on
The cited links in those answers show where your category’s authority sits, so log the domains. In our experience, the same handful comes up again and again for US small businesses: a directory or two, a review platform, a trade association, a local publication, one or two Reddit threads.
That list is your off-page to-do list, in priority order. If four of your ten prompts cite the same directory and your listing there is wrong or missing, fixing it is worth more than another blog post. The list also shows which sources you can never control, and where you need reviews rather than pages.
What Search Console gives you for free
On Google’s side, you no longer have to guess. Since August 31, 2026, the generative AI performance report has been available to all websites worldwide. It shows impressions from AI Overviews and AI Mode by page, country, date and device. Before you read the chart, note how it aggregates: if two of your pages appear in the same AI feature, that counts as one impression in the chart total.
Check one setting while you’re in there. Search Console’s Search generative AI control under Settings decides whether your site is eligible at all; if a previous developer set it to exclude, Google says you “won’t receive any traffic or impressions from these features”. The equivalents elsewhere are robots.txt rules. OpenAI states that “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links” (Overview of OpenAI Crawlers), and Perplexity asks you to allow PerplexityBot and its published IP ranges through your firewall. From the outside, an accidental block looks exactly like an AI visibility problem.
What the report doesn’t give you, at least as of September 2026, is the wording of the answer, the prompt behind it, or anything outside Google. That’s why you still need the manual sheet.
When a paid AI search monitoring tool earns its keep
A spreadsheet is enough when you have one location, one service line and ten prompts; at that size the monthly run is a small job you can do yourself. A tool starts to pay when you have several locations or brands to compare, when you want sentiment and share of answers tracked daily rather than monthly, or when somebody other than you has to read the report.
Be skeptical of the metrics. “Visibility score” means something different in every product, and none of them can see inside ChatGPT any more than you can. They run prompts on a schedule, the same as your sheet, only with a larger sample. What you’re paying for is that sample size and the reporting.
Common questions
How often should I run the test?
Monthly is enough for most small businesses, on the same day each month. Running it weekly mostly tells you how noisy these systems are.
Why do I get a different answer every time I ask?
These models are non-deterministic, and they personalize answers based on context and history. That’s why the method uses a fixed prompt list, a fresh session and a trend line instead of a single screenshot.
Is being mentioned worth anything if nobody clicks?
Often, yes, because it works like a recommendation. Google reports that clicks arriving from results pages with AI Overviews are “higher quality” (AI features and your website). Judge it on calls and inquiries over a quarter.
Is this the same thing as SEO reporting?
It’s a separate layer of measurement on top of much of the same work. We explain what differs in generative engine optimization vs traditional SEO.
What to do when you’re not in the answers
After three months of the sheet, you’ll know which of two problems you have. If you’re never cited, it’s a content and eligibility problem, and the fix is the sequence in how to rank in AI Overviews. If you’re cited but never recommended, it’s a reputation problem, which means working on reviews, directory accuracy and third-party mentions.
If you’d rather have the baseline run for you, our AEO services start with an AI visibility audit. Send us your ten prompts and we’ll tell you which businesses the answers are naming instead of you.
