# What Is AI Visibility and How Is It Measured?

> What AI visibility is, why there is no standard formula, what data Google and Bing give you and how to tell whether assistants name your products.

URL: https://clione.ai/glossary/ai-visibility/
Lang: en

"AI visibility" is a phrase from tool dashboards and agency proposals, and it almost always arrives with a number next to it. Behind it is a question any store understands: **when a customer asks an assistant to recommend a product in your category, do you show up?** It is not the SEO visibility of old, which is calculated from positions in a list of results, or the traffic that reaches your site. It is presence inside the answer. And you should know from the start that it has no single definition: each tool, each study and each search engine counts it its own way, so the figure only makes sense next to its method.

## What it includes

A brand can be present in a generated answer in three different ways, and each is counted in a different unit:

- **As a linked source.** The answer carries a link to one of your pages next to the sentence it backs: an [AI citation](/glossary/ai-citation/). It can be clicked and can end in a visit.
- **As a named mention.** The answer names your brand or your product without linking any page. It can weigh on the reader's decision, but it leaves no trace in your analytics.
- **As a link shown in Google's AI features.** Google counts a [search impression](/glossary/search-impression/) each time a link to your site is shown in [AI Overviews](/glossary/ai-overviews/) or in [AI Mode](/glossary/google-ai-mode/).[^gsc-genai-report]

**What the search engines give you, first-hand.** The [generative AI performance report](/glossary/generative-ai-performance-report/) in [Search Console](/glossary/google-search-console/) gives your site's impressions in AI Overviews and in AI Mode, by page, country, device and date, without clicks or queries. Google opened it to all sites on August 31, 2026, and it only appears if the site accumulates enough impressions in those features and is not excluded from them.[^gsc-genai-report] The AI Performance report in [Bing Webmaster Tools](/glossary/bing-webmaster-tools/), in public preview since February 10, 2026, counts citations: how many times your content appears as a source in [Copilot](/glossary/copilot/), in Bing's AI-generated summaries and in some partner integrations, how many of your pages are cited on average per day, the phrases the AI used to retrieve your content, the citations for each URL and how they change over time. Bing presents it with an idea that sums the term up: **visibility is no longer just about blue links**.[^bing-ai-performance]

**What you measure by asking.** In ChatGPT, [Gemini](/glossary/gemini/), [Perplexity](/glossary/perplexity/) or Claude, visibility is measured by running queries and reading what they answer, which is what tracking tools do. That measure is a sample that someone designs: which queries, in which assistants, how many times, what counts as presence. Google adds a warning that holds for any dashboard: **no third-party tool has access to its internal ranking or AI systems**, and it asks you to be wary of those that promise rankings or claim to use internal Google metrics.[^google-ai-guide]

**Why there is no formula.** Not even the academic work that gave [GEO (Generative Engine Optimization)](/glossary/geo/) its name takes one for granted: it presents itself as a framework for optimizing content visibility in [generative engine](/glossary/generative-engine/) responses **and for defining their visibility metrics**. Its most quoted figure, that GEO can boost visibility by up to 40%, is measured with the metrics the paper itself defines.[^geo-paper] Google, for its part, describes [AEO (Answer Engine Optimization)](/glossary/aeo/) and GEO as the work centered on improving visibility in [AI search](/glossary/ai-search/) experiences, and assigns no measure to it.[^google-ai-guide]

## In a store

A backpack store gathers in one table, for a single [product page](/glossary/product-page/) and a single month, everything it can know about its AI visibility. It is a **fictional example**: the product page and the figures are made up to show that each source counts a different thing.

| Source (fictional example)                 | What it counts                               | Urban backpack 20 L |
| ------------------------------------------ | -------------------------------------------- | ------------------- |
| Search Console generative AI report        | Link impressions in AI Overviews and AI Mode | 210                 |
| Bing Webmaster Tools AI Performance        | Citations of the URL as a source             | 14                  |
| Own panel in ChatGPT: 10 queries × 3 times | Answers that name the product                | 7 of 30             |
| Own panel in ChatGPT: the same 30 answers  | Answers that link the product page           | 2 of 30             |
| Own panel in Gemini: 10 queries × 3 times  | Answers that name the product                | 4 of 30             |


Five rows, four units: impressions, citations, mentions and links. **They are not added or averaged**, and any "visibility index" that merges them into one number has made decisions worth knowing about. What the table does say is useful: the product shows up in ChatGPT far more than it is linked, so many of those appearances cannot bring a visit. And it is built by product, not by brand: a store gets more from knowing which product pages assistants name than from knowing how many times its name comes up.


**Appearing in an answer is not getting visits.** Bing warns that its figures say how often your content is used as a source, but **do not indicate the ranking, the authority or the role of a page within a particular answer**,[^bing-ai-performance] and Google's report gives impressions, not clicks.[^gsc-genai-report] The visits that presence brings, when it brings them, are a different measure, [AI referral traffic](/glossary/ai-referral-traffic/), and are read in your analytics.


## How to check it

1. **Open the Search Console generative AI performance report** and filter by your most important product pages and categories. If it does not appear, the site may not yet accumulate enough impressions or may be excluded from those features.[^gsc-genai-report]
2. **If your site is verified in Bing Webmaster Tools, open AI Performance** and look at which URLs are cited and with which retrieval phrases.[^bing-ai-performance]
3. **Set up a fixed panel of purchase queries**, without your brand, and always run them the same way in ChatGPT and in Gemini, in new chats and on several days. Note the named products and the cited URLs separately. Clione asks this question for you: it asks ChatGPT and Gemini what shoppers ask and shows whether they name the store's products, whether they keep recommending it when the shopper asks them to compare and pick one, and who takes its place when they stop.
4. **When a tool or an agency gives you a figure, ask for the method**: the queries, the assistants, the repetitions, the dates and what counts as presence. Without that, the number cannot be compared with any other.
5. **Cross-check presence against visits.** If the answers name you and your analytics shows no visits from assistants, the presence is mostly unlinked mention.

## Related concept: AI share of voice

[AI share of voice](/glossary/ai-share-of-voice/) starts from the same presence and sets it against your competitors': of all the space brands take up in the answers of your market, what share is yours. AI visibility looks at it on its own: whether you appear, where and how. The two can move in opposite directions. If the answers in your category start naming more brands, you can show up more often than before and still hold a smaller share of the total. **Visibility says whether you are there; share of voice, how much of the space is yours.**

To keep going, the [measurement and verification reading path](/learn/measurement-and-verification/) orders the rest of the terms, and the entry on [attribution of visits that arrive from AI](/glossary/llm-attribution/) explains how to tell whether that presence ends in orders.