How do you measure brand visibility in AI search?
Measure it in three layers. First, run a fixed set of prompts your buyers actually ask through ChatGPT, Gemini, Perplexity and Google AI Mode on a schedule, and count how many answers mention your brand or cite your site. Second, use the AI data in Search Console and Bing Webmaster Tools. Third, track AI-referred traffic in GA4.
Each layer answers a different question. The prompt set tells you whether you're mentioned at all. Platform reports tell you which of your pages are used as sources. GA4 tells you whether that visibility sends visitors. Looking at only one of them gives you a skewed picture.
Why measure visibility separately from traffic?
AI answers often resolve the question without a click. A shopper asks "which brand for women's running shoes", sees your brand in the answer and a few days later searches your name on Google. GA4 records that visit as branded organic search, and the AI's part in it disappears.
If you look only at traffic, you'll underestimate the effect. We explain how to separate the traffic side in our guide on how to track ChatGPT and AI traffic in GA4. This post focuses on the step before the click: how often your brand shows up inside the answers.
Which metrics should you track?
| Metric | What it tells you | Where to measure it |
|---|---|---|
| Mention rate | Share of answers in your prompt set that name your brand | Prompt set tests, by hand or with a tool |
| Citation rate | Share of answers that link one of your pages as a source | Prompt set tests, Bing AI Performance |
| Position | Whether you're named first in lists or near the end | Prompt set tests |
| Competitor share | How often competitors are named for the same prompts | Prompt set tests |
| Accuracy and tone | Whether the answer says correct, favourable things about you | Prompt set tests, manual review |
| AI impressions | Impressions of your pages in Google's generative AI features | Search Console (where the report is available) |
| AI-referred sessions and revenue | Visits and sales the visibility brings to the site | GA4 |
The first three metrics are enough to start. Don't skip accuracy: if an answer shows a wrong price, an expired promotion or an old returns policy, that's more urgent than visibility.
How do you build a prompt set?
The quality of the measurement depends on the quality of the prompts. Build the set like this:
- Collect purchase-intent questions. Customer service tickets, long queries in Search Console and objections your sales team hears are good sources.
- Split them into three groups. Category prompts ("most durable stroller"), comparison prompts ("X or Y") and brand prompts ("is brand X reliable").
- Start with 20 to 50 prompts. Fewer leaves too much to chance; many more becomes unmanageable by hand.
- Run each prompt several times. Answers can change from run to run, so ask each one at least three times and average.
- Keep conditions fixed. Test logged out or with a fresh account, in the same language and location setting. Note the test date and model version.
- Log results in a table. For each answer: were you mentioned, were you cited, in what position, which competitors appeared, was the information correct.
- Repeat monthly. Weekly runs usually show more noise than signal.
Example: run a 30-prompt set three times and you have 90 answers. If your brand appears in 18 of them, your mention rate is 20 percent. Tracking that next month alongside your competitors' rates means far more than a single number on its own.
Write prompts in the language your buyers use. If you sell in Turkey, test Turkish prompts phrased the way people talk, such as "which brand would you recommend" or "affordable but durable". Don't generalise results from English prompts to the Turkish market.
What do Search Console and Bing Webmaster Tools show?
Google's documentation says sites appearing in AI Overviews and AI Mode are included in the overall traffic in the Search Console Performance report, under the Web search type. So for a long time this data couldn't be separated. In June 2026 Google announced generative AI performance reports for Search Console and is rolling them out gradually to a subset of sites. If you have the report, you can see impressions in AI features by page, country and device; if not, you'll need to keep working with the standard Performance report.
On the Microsoft side, Bing Webmaster Tools opened its AI Performance report in public preview in February 2026. According to Microsoft's announcement, it shows total citations, the average number of cited pages per day, the grounding queries the AI used when retrieving content, and citation counts per page. Coverage is limited to Microsoft Copilot, AI-generated summaries in Bing and select partner integrations. There's no comparable official report for ChatGPT, Gemini or Perplexity yet, which is why prompt set testing is still needed.
Are paid AI visibility tools worth it?
A wave of paid tools now tracks AI visibility. The upside is clear: they run your prompt set automatically, keep history and chart competitor comparisons. There are downsides too. Many collect answers through APIs, and those can differ from what a user sees in the app. And the more prompts a tool runs, the more it costs.
The healthy route we see in practice: run the first two or three months by hand or with a simple spreadsheet. Once you know which prompts matter to you, pick a tool. When choosing, ask which models it covers, how it collects answers and whether you can export your data.
Why do AI answers change every time?
Language models don't give the exact same answer to the same prompt every time. Models that search the web pick different sources depending on the live results. The user's location, language, history and even the wording of the prompt change the outcome. That's why concluding "ChatGPT recommends us" or "it never mentions us" from one screenshot is a mistake. Repeated runs and averages are the only practical way to manage the noise.
How do you turn measurement into action?
Once you have data, look at three questions. For which prompts are competitors named while you aren't? Do you have a page that answers those prompts, and is it clear and current? Which outside sources do the answers cite? Comparison articles, forums and news sites often come up as sources, and being represented accurately on them matters as much as updating your own site. We cover the content side in our GEO guide to getting cited in ChatGPT and Perplexity.
If you'd like AI visibility handled together with SEO, see our ecommerce SEO and GEO service. To see where your brand stands today with the Performetic team, you can request a free growth audit.
Key takeaways
- AI visibility is measured in three layers: prompt set tests, platform reports and GA4 traffic.
- Mention rate, citation rate and competitor share are enough to start; check accuracy too.
- Pick 20 to 50 purchase-intent prompts, run each several times and repeat monthly.
- Search Console's new AI reports are rolling out gradually; Bing AI Performance shows citations in Copilot and Bing summaries.
- Learn which prompts matter through manual measurement before paying for a tool.
