AI visibility is how often, and how accurately, AI assistants mention your brand when people ask questions in your category. You measure it by running a fixed set of prompts many times and tracking how often you appear.
Somewhere right now, someone is asking an assistant which agency to hire, which tool to buy or which company to trust. The answer arrives as a short list and a confident paragraph. No results page. No second page to scroll to. Possibly no click.
Most companies have no idea what those answers say about them. This article explains how to find out, what is worth tracking and which numbers to distrust.
What AI visibility includes
Being mentioned is only the first layer. A useful picture has four parts:
- Presence. Does your brand appear at all when someone asks about your category?
- Accuracy. Is what the AI says about you correct and current?
- Framing. Are you described the way you would want to be, and for the right customer?
- Sources. Which websites is the answer drawing on when it talks about you?
The second and third parts are where the real brand risk sits. An assistant that recommends you for the wrong thing, or describes the company you were three years ago, is sending you the wrong customers.
Why “AI ranking” is mostly a mirage
Here is the fact that should shape your entire approach: AI answers are not stable.
In a study published in January 2026, SparkToro and Gumshoe.ai had 600 volunteers run 12 recommendation prompts through ChatGPT, Claude and Google’s AI features, 2,961 times in total. They reported a less than 1 in 100 chance that ChatGPT or Google’s AI would give the same list of brands in any two responses, with Claude only slightly more likely to repeat itself. Getting the same list in the same order was rarer still.
Their conclusion is the useful part. Claiming a “ranking position” in AI does not hold up, but visibility percentage across many prompts, run many times, is a reasonable metric.
The authors are open about the limits. The work was not peer-reviewed, conditions were not controlled, the data comes from late 2025, and one co-author works at an AI tracking company. Treat it as a strong signal, not settled science. But it is enough to be suspicious of any report telling you that you are “number three in ChatGPT”.
How to run an AI visibility audit
You can do a first version in an afternoon with a spreadsheet:
- Write 20 to 30 prompts. Use the language customers actually use. Pull it from sales calls and support tickets, not from your keyword list.
- Cover different kinds of intent. Category questions (“best branding studios in Tel Aviv”), problems (“why is our website not converting”), comparisons (“X vs Y”) and direct questions about your company.
- Test several assistants. Each product behaves differently, so check the ones your customers use.
- Repeat every prompt. One run tells you almost nothing. Run each several times, in fresh sessions, and more if you can.
- Record what matters. Whether you appeared, who appeared alongside you, what was said about you and which sources were cited.
- Calculate your visibility rate. The share of runs in which your brand appeared, overall and for each group of prompts.
- Do it again on a schedule. Monthly or quarterly is plenty. You are watching a trend, not a scoreboard.
One thing to keep in mind while testing: assistants can personalize based on memory and location. Use clean sessions where you can, and do not panic over a single answer on your own account.
What to track over time
A simple dashboard beats a sophisticated one nobody opens. Keep it to these:
- Visibility rate for each prompt group.
- Share of voice: how often you appear compared with the competitors who show up alongside you.
- Error log: every wrong claim about your pricing, services, location or history.
- Top cited sources: the sites assistants lean on in your category.
- AI referral traffic: visits from assistants in your analytics, kept as its own segment.
- Branded search and direct inquiries: people often hear a name in one place and look it up in another.
Be careful with attribution. Google’s documentation says traffic from AI Overviews and AI Mode is counted within the overall Web search data in Search Console rather than reported separately, so you cannot simply pull a clean AI number from there. And an answer that shapes a decision may never produce a click at all. Asking new customers how they heard about you is low-tech and still very useful.
How to improve what AI says about you
The audit will usually point to one of three problems, and each has a different fix.
You do not appear at all
Check that the right crawlers can reach your site, then look at the sources being cited for your category. If you are absent from them, that is your distribution plan: earned press, credible directories, reviews and expert commentary. It helps to have content that actually answers the prompts, as covered in our guide to GEO.
You appear, but the details are wrong
Fix the facts at the source. Make your own site unambiguous about what you do, where you are and who you serve. Then correct outdated profiles, directory listings and old articles. Answers often lag behind reality, so expect this to take time.
You appear, but sound like everyone else
This is the most common problem, and the hardest. If your company describes itself in generic language, the model has nothing distinctive to repeat. The solution is not a technical one. It is sharper positioning, stated the same way in every place you control.
The test
Ask three assistants to describe your company in two sentences. Then ask who your closest competitors are and why someone would pick you over them.
Read the answers the way a prospective customer would. If the description is vague, you have a message problem. If it is wrong, you have a facts problem. If you are missing altogether, you have a distribution problem.
All three are fixable. You just have to look first. If you want help with the part that comes after the audit, talk to us.
Frequently asked questions
What is AI visibility?
AI visibility describes how often and how accurately AI assistants such as ChatGPT, Claude and Google’s AI features mention a brand in their answers. It includes whether the brand appears, how it is described and which sources the answer relies on.
How do you measure AI visibility?
Run a fixed set of realistic customer prompts across several AI assistants, repeat each prompt multiple times and record how often the brand appears. The result is a visibility percentage that can be tracked over time and compared with competitors.
Can you track your ranking in ChatGPT?
Not reliably. A 2026 SparkToro study found that AI tools rarely return the same list of brands twice, so a fixed position has little meaning. The authors concluded that visibility percentage across many prompts and repeated runs is a more reasonable metric.
Does Google Search Console show AI Overview traffic?
Google says traffic from AI Overviews and AI Mode is included in the overall Web search data in Search Console. It is not listed as a separate report, so it cannot be isolated directly from those totals.
How can a brand correct wrong information in AI answers?
Correct the facts on your own website first, then update outdated profiles, directories and third-party articles that AI systems may cite. Consistent, accurate information across trusted sources is the most practical lever, though answers may take time to change.
