AI visibility is whether customers hear your brand's name when they ask an assistant a question, and whether the assistant points them to your pages. It is not the same as rankings, and it is not the same as clicks. A citation can shape a buying decision without producing a click, so a rankings report alone cannot tell you whether customers hear your name. Measuring AI visibility means measuring three things separately: whether your pages appear inside AI answers, whether assistants name your brand when asked, and whether those mentions turn into visits.
No vendor publishes a visibility score, a formula, or a benchmark for AI visibility. None is claimed here. What follows is a measurement routine built from the reports vendors do publish and from asking the assistants directly, with the limits of each method stated plainly.
1. Start with what Google measures
Google is the only engine with a report that shows your pages inside its AI features. Use it first, because it is the only measurement in this guide that comes from the vendor itself.
- Open Search Console and find the Generative AI performance report. Google's AI guide points site owners to this report to see how content performs in generative AI features on Google Search and Discover. Check the vendor docs in Sources for the current path to the report.
- The report shows one metric: impressions. An impression is how many times links to your site were shown to a user in a generative AI feature on Google Search, which currently means AI Overviews and AI Mode. A link being shown is not a click. The user does not have to click it, scroll to it, or even notice it.
- The report breaks impressions down by page, country, device, and date. It does not break them down by query. You can see that a page was shown inside an AI answer. You cannot see what was asked.
- The report has no clicks, no click-through rate, and no average position. Clicks from AI Overviews and AI Mode do exist, and they sit inside the ordinary Performance report under the Web search type, mixed in with every other organic click, so you cannot pull them out on their own. Do not divide overall clicks by AI-only impressions and call the result a click-through rate for AI features. That number is not a real rate, because the clicks and the impressions come from different sets.
- The same impressions are also counted in the main Performance report. Do not add the Generative AI impressions to your Search impressions and call the result a total. The report filters impressions you already had, it does not count a new pool of exposure.
- A missing report, or a report with no data, does not prove you have no AI visibility. It can mean too few impressions to show, or that the site is excluded from Search generative AI features. Check the setting in Search Console before concluding anything.
2. Ask the assistants directly
For ChatGPT, Gemini, Perplexity, Claude, and Copilot, there is no vendor report that shows your brand. The measurement is asking them directly with a fixed set of prompts and logging what they say. That is the only way to see it, and it has to be done the same way every time or the numbers cannot be compared.
- Write a fixed prompt set before you start. Use the questions your buyers actually ask, not questions about your brand. A buyer asks for the best trail shoe for wet weather, or a coffee grinder for espresso under a budget, or a gift for a new parent. Write twenty to thirty of those questions, covering your main products, your main categories, and your main buyer problems.
- Add a small set of branded prompts: questions that name your brand directly, and questions that ask for alternatives to your brand. The branded set tells you whether the assistant knows you at all. The unbranded set tells you whether it recommends you when it has a choice.
- Run the same prompts, in the same order, on the same assistants, on a fixed schedule. Monthly is enough for most stores. Weekly is better when you are launching a product or fixing a problem. Do not run the prompts once and judge from one run, because answers change from run to run.
- Log the same fields every time: the assistant, the prompt, the date, whether your brand was named, whether your brand was recommended first or named among others, which pages the assistant cited, and whether the facts it stated about your products were right or wrong. A log with those fields is a measurement. A memory of what you saw last time is not.
- Keep the prompts in a file the whole team can see, and do not change them between runs unless the business changed. Changing the prompts changes the measurement, and you will not know whether the brand moved or the questions moved.
3. Watch the visits that follow
A mention that brings no visit is still a mention, but a mention that brings visits is the one that pays. Watch the visits with the analytics the store already has.
- In the store's analytics, look for referral traffic from the assistants that send it. Some assistants send a referrer the analytics can see, and some do not. Do not treat a missing referrer as proof of no traffic, because a visit can arrive with no referrer at all.
- Watch the pages that the Generative AI report says were shown, and see whether their visits move in the same direction. A page shown more often in AI answers, with visits flat, is a page that is seen but not clicked. A page shown more often, with visits rising, is a page that is seen and clicked.
- Watch branded search separately from unbranded search. A rise in branded search after a period of AI mentions can mean customers heard the name from an assistant and then searched for it directly. That is a real effect, and it will not show up as a referral from an assistant.
- Do not expect a clean line from a mention to a sale. A customer can hear a brand from an assistant, search for it later, visit the site, and buy days after that. The measurement in this guide is visibility, not attribution, and it should not pretend to be attribution.
4. Track the facts, not just the name
A brand can be named with wrong facts, and that is worse than not being named at all. A wrong price, a discontinued product recommended as in stock, or a shipping claim the store cannot meet will cost a sale and teach the assistant not to trust the brand.
- In the prompt log, add a column for factual accuracy. For every mention, note whether the price, the availability, and the key product facts the assistant stated were right.
- When a fact is wrong, trace it to the source. Check the page, the markup, and the feed for that product, and check whether the assistant cited a page that is out of date. A wrong fact usually comes from a stale page or a mismatch between the three copies of the truth, not from the assistant inventing it.
- Fix the source, then retest with the same prompts after the fix has been recrawled. Google's AI guide says its systems retrieve relevant, up-to-date pages from the Search index, so a corrected page that has not been recrawled is still the old page to the system that retrieves it. Ask for a recrawl in Search Console after a real fix, and check the vendor docs in Sources for the current path.
- Count wrong facts separately from missing mentions. A brand with many mentions and many wrong facts has a different problem from a brand with no mentions at all, and the fix is different too.
5. What a good measurement looks like
A good measurement is a small set of numbers, tracked the same way, over time. It is not a single score, because no vendor publishes one, and it is not a dashboard, because no single dashboard shows all of it.
- Google AI impressions, by page, from the Generative AI performance report, tracked monthly.
- Share of prompts where your brand was named, by assistant, from your own prompt log, tracked monthly.
- Share of prompts where your brand was recommended first, by assistant, from the same log.
- Share of mentions with wrong facts, by assistant, from the same log.
- Visits to the pages the report says were shown, from the store's analytics, tracked monthly.
- Branded search volume, from Search Console, tracked monthly.
Six numbers, tracked the same way, over time, will tell you more than a single score that no vendor publishes. The direction matters more than the level. A share rising from one in ten prompts to three in ten is progress, even if the absolute number is still small. A share falling while the store's pages have not changed is a signal to check what changed on the other side: a competitor's pages, a crawl problem, or a stale fact.
The fundamentals
These mechanics are shared across every engine and covered in full in their own guides:
- •The llms.txt format and a complete sample, see our guide to writing a good llms.txt file
- •The markup this engine reads, see our product schema guide
- •The complete crawler rule reference, see our robots.txt guide for AI search
Say you run a pet food brand
Say you sell pet food, and you start measuring for the first time. The Generative AI report shows your sensitive stomach guide was shown in AI Overviews, but your product pages were not. The prompt log shows ChatGPT names your brand in three of twenty prompts, and recommends it first in one. Perplexity names it in five, and cites the guide twice. Claude does not name it at all. The facts check finds one wrong fact: ChatGPT states a price from before a sale ended, because the feed still shows the old price. You fix the feed, ask for a recrawl, and run the same prompts a month later. The share rises, the wrong fact is gone, and the guide's visits rise with it. You did not buy a visibility score. You asked the same questions, logged the same fields, and watched the direction.
What this guide does not cover
This guide covers how to measure AI visibility. It does not cover how to track mentions over time in detail, which is our guide to tracking AI mentions of your brand over time, or how to run a full audit, which is our guide to running an AI visibility audit on your store, or how to benchmark against competitors, which is our guide to benchmarking AI visibility against competitors. Measurement tells you where you stand. The other guides tell you what to do about it.
Check your own brand now
Run the free scan. It takes seconds, no signup, and shows whether AI crawlers can read your pages today: aizii.ai/aeo-scan
Run the free scanSources
- Google Search Central, Optimizing your website for generative AI features on Google Search (specific): https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central, AI features and your website (specific): https://developers.google.com/search/docs/appearance/ai-features
- Google Search Central, Creating helpful, reliable, people-first content (specific): https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- OpenAI, Bots documentation (specific): https://platform.openai.com/docs/bots
- Perplexity, Bots documentation (specific): https://docs.perplexity.ai/guides/bots
Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
