AI Visibility: How to Measure and Improve Your Brand in AI Answers

What AI visibility means

AI visibility is how often AI assistants such as ChatGPT, Claude, Perplexity, Gemini, Microsoft Copilot and Google's AI Overviews and AI Mode mention or cite your brand when people ask questions you should show up for. You measure it by asking a fixed set of buyer prompts every month and logging mentions and cited sources, and by reading the AI reports in Google Search Console and Bing Webmaster Tools. You improve it with crawler access, clear facts about your brand, and being described consistently on the third-party pages assistants read.

It covers two different things:

• Mentions: the assistant names your brand in its answer, for example as one of the tools it recommends.
• Citations: the assistant links to one of your pages as a source.

You can be mentioned without being cited (the model already knows you, or it cites a review site that lists you) and cited without being recommended (your guide answers a how-to question). Track both.

Each assistant draws on two places: what its model learned in training, and pages it retrieves live when it searches. What a model learned in training changes only when its maker ships a new model. Retrieval can change as soon as new pages are indexed and chosen as sources, and that is where most of your work has an effect.

Where each assistant gets its sources

Knowing which crawler and index feed which assistant tells you what to check. From each vendor's own documentation:

• ChatGPT search uses OpenAI's OAI-SearchBot. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. GPTBot is a separate crawler for model training.
• Perplexity uses PerplexityBot to surface and link sites in its answers, and says it is not used to crawl content for training foundation models. Perplexity-User fetches pages when a user asks a question.
• Claude uses Claude-SearchBot to improve its search results and Claude-User to fetch pages for a user's question. ClaudeBot collects content that may be used for training.
• Google AI Overviews and AI Mode draw on Google Search. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link, and that there are no additional requirements or special optimizations.
• The Gemini app can ground answers in content from the Google Search index. The Google-Extended token controls whether your content is used for Gemini training and for that grounding, and Google says it does not affect inclusion or ranking in Google Search.
• Microsoft Copilot citations are counted in Bing Webmaster Tools' AI Performance report, so being well indexed in Bing matters.

Our guides to getting recommended by ChatGPT and appearing in Google AI Overviews go deeper on each.

How to check if ChatGPT recommends your brand

The most dependable measurement is still manual, and it takes about 30 minutes a month:

1. Write 20 to 30 prompts your buyers would ask, in three groups: recommendation prompts ("best tool for X for small teams"), problem prompts ("how do I fix Y") and brand prompts ("what is [your brand]", "[your brand] pricing", "[your brand] vs [competitor]").
2. Run every prompt in each assistant you care about: ChatGPT with search on, Perplexity, Gemini, Claude, Copilot and Google AI Mode. Use a logged-out or fresh session so past chats do not shape the answer.
3. Log the result in a sheet with one row per prompt and assistant: mentioned (yes or no), position in any list, whether the description is accurate, the URLs cited and the competitors named.
4. Repeat on the same day each month with the same prompts. Answers vary from run to run, so read the trend across the whole set, not a single answer.

The cited URLs are the most useful column. They show which pages each assistant relies on for your topic: the review sites, comparison articles, directories and threads you should be on next.

Brand prompts test whether assistants know who you are. For a new brand, expect thin or missing answers at first. They improve as your profiles, listings and mentions get indexed, which is why the work in the next sections matters more than the measurement.

Free AI reports from Google and Bing

Two first-party reports now show part of the picture for free:

• Search Console's Generative AI performance report shows how often links to your site appeared in Google's AI Overviews and AI Mode, by page, country, device and date. Google launched it in June 2026 and had rolled it out to all websites by 31 August 2026. It reports impressions; traffic from AI features is also included in the main Performance report under the Web search type. The report may not appear if your site has very few AI impressions.
• Bing Webmaster Tools' AI Performance report, in public preview since February 2026, counts how often your pages are cited in Microsoft Copilot, Bing's AI-generated summaries and some partner integrations, with your cited pages and a sample of grounding queries. Bing notes that citation counts do not indicate placement, ranking or authority.

Check one setting while you are there. Search Console has a Search generative AI control in Settings. Sites are included by default; a site that has been opted out gets no impressions or traffic from these features.

Google's own advice on measurement is worth repeating: use these reports, and be sceptical of third-party tools that claim access to Google's internal metrics.

AI visibility tools: what they do and what to check

AI visibility tools automate the manual method: they run your prompts across assistants on a schedule and chart mentions, citations and competitors. They save time once you track more than a few dozen prompts or several competitors. What they do not do is change the answers. That takes the work in the next four sections.

Before you pay for one, check:

• Which assistants it covers. Some entry plans cover a few and charge more for the rest.
• How many prompts, and how often they run. A useful benchmark needs the 20 to 50 questions your buyers really ask.
• Whether it stores cited sources, not only mentions. Sources tell you where to act.
• Competitor benchmarking and export, so the data is not trapped in the tool.

We compare trackers and services, with prices checked in October 2026, in the best GEO tools and agencies. For individual tools, see our comparisons of Otterly.ai, Peec AI and Profound.

If you are just starting, a spreadsheet is enough. Buy a tool when the manual run takes longer than the work it informs.

Lever 1: let AI crawlers reach your pages

An assistant cannot cite a page it is not allowed to fetch. Check your robots.txt for each search crawler above, and decide separately about training crawlers:

• Allow for visibility: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot.
• Your choice: GPTBot, ClaudeBot and Google-Extended govern training (and, for Google-Extended, grounding in the Gemini app).

These are separate robots.txt tokens, so you can block the training crawlers and still allow the search ones. Google states that Google-Extended does not affect inclusion in Google Search.

Check two more places. Page-level robots meta tags such as noindex and nosnippet limit what Google can show, including in its AI features. And some CDN and firewall settings can block AI crawlers before they ever read your robots.txt, so test the live site rather than the file alone.

Our free AI crawler checker reads a page's robots.txt rules, robots meta tags and llms.txt and shows which AI crawlers can reach it. If the file needs fixing, the robots.txt generator has presets for the common AI crawlers.

Lever 2: make your brand easy to describe

Assistants describe you with whatever text they find, so give them clear, consistent facts:

• An About page that answers the basics in plain sentences: what you do, who it is for, what it costs, when you were founded, where you are based and what you are not.
• The same name and one-liner on every profile and listing. Our guide to social profile backlinks has a facts sheet you can reuse.
• Organization structured data with your name, logo and sameAs profile links. Google says structured data is not required for its AI features and that there is no special schema to add, but it is still the cleanest way to state official facts. Our schema markup generator builds it.
• An llms.txt file, if you want one. It is a proposal from September 2024 for a Markdown map of your key pages. Google says Google Search does not use llms.txt, and we have not found documentation from other assistant makers saying they use it to choose sources, so treat it as cheap hygiene that is most useful for developer docs. Draft one with the llms.txt generator, validate it with the llms.txt checker, and read our llms.txt guide for the format.

If your name is similar to another company's, say clearly what you are not. Assistants that confuse two companies usually do so because nothing on the web separates them.

Lever 3: be described on the pages assistants read

Your own site is one source. Assistants that search the web also read what other sites say about you, and recommendations usually come from there. Use the cited URLs from your prompt log to pick targets:

• Comparison and "best X" articles that rank for your category. Ask the author for an update when you are a real fit, with facts they can check.
• Review and software directories such as G2 and Capterra, and the launch directories your buyers browse.
• Reddit and community threads. Threads often rank in Google, and assistants can cite them. Contribute helpful, disclosed answers; never fake reviews or upvotes. See Reddit marketing for SaaS.
• Podcasts, newsletters and guest articles where you explain something specific.
• Original data. A survey or analysis gives writers a reason to cite you, and their articles become pages assistants read.

Keep the description consistent everywhere: the same name, one-liner and category. Inconsistent facts across third-party pages make it harder for an assistant to describe you correctly. Our guide to GEO covers this work in more depth.

Lever 4: publish pages that answer the prompts

For each prompt where a competitor gets mentioned and you do not, ask whether your site has a page that answers it directly. The missing pages are usually:

• Comparison pages: you versus a named competitor, fair and specific.
• Alternatives pages: "alternatives to [competitor]", including when the competitor is the better fit.
• Use-case pages: who the product is for, by role or industry, with real examples.
• Pricing in plain text, not only in an image or behind a form.
• How-to guides for the problems your product solves.

Open each with a direct two- or three-sentence answer, then the detail. Google's guidance for its AI features is to write unique, people-first content with first-hand experience. It says there is no need to split content into tiny chunks for AI, and it warns that mass-producing pages for query variations can break its scaled content abuse policy. Quality beats volume here too.

Link the new pages from your home page or main navigation so they are crawled quickly, and add them to your sitemap.

A 30-day plan to improve AI visibility

• Week 1, baseline: write your prompt set, run it in every assistant and log the results. Open the Generative AI report in Search Console and AI Performance in Bing Webmaster Tools. Run the AI crawler checker on your home page and two key pages.
• Week 2, access and facts: unblock any search crawlers, rewrite the About page around plain facts, add Organization schema with sameAs, and publish llms.txt if you want one.
• Week 3, third-party mentions: from your log, list the five sources assistants cite most for your category. Get listed or mentioned on two of them, complete your key profiles, and post one genuinely helpful community answer.
• Week 4, answer pages: publish two pages that answer prompts you lost, such as a comparison and a use-case page, and link them from your home page or main navigation.
• Day 30: rerun the prompt set and compare it with week 1.

Be patient with the result. Answers that depend on live search can change once new pages and mentions are indexed and chosen as sources; what a model learned in training changes only with a new model. No one can guarantee a mention in ChatGPT or any other assistant, and anyone who promises one is guessing.

If you would rather hand the work off, our GEO service benchmarks your buyer prompts, fixes crawler access, schema and llms.txt, builds citations, and re-tests every month with the results in your dashboard.