AI Visibility Audit: How to Check What AI Assistants Say About Your Brand
An AI visibility audit is a structured read of what AI assistants such as ChatGPT, Gemini, Copilot, Perplexity and Google’s AI Overviews say about a brand or a person, and which sources they rely on. It grades the result in three pillars, Sentiment, Discoverability and Authority, and ends in a prioritized list of gaps to fix.
What is an AI visibility audit?
Buyers, investors, partners and candidates now ask an assistant before they open a website. Who is this company? Can it be trusted? Which provider is best for my case? The answer they get is a summary the assistant writes from the sources it trusts, and the brand has no direct say in it.
An AI visibility audit records those answers under controlled conditions and grades them. Three questions sit behind every grade:
- Sentiment: how the answers describe you, in their own words.
- Discoverability: whether you are named when buyers ask the questions that matter, and how prominently.
- Authority: whether the sources behind the answers are strong, current and accurate.
Any brand or person that people ask AI about by name needs one: companies in regulated or high-trust sectors, founders and executives, and online businesses whose buyers now build their shortlist in a chat window.
What an AI visibility audit reads: the AI engines and the sources they cite
An audit reads the answers and the evidence behind them. The answers come from several AI engines side by side, among them ChatGPT, Gemini, Copilot, Perplexity and Google’s AI Overviews. Each can describe the same company differently, because each draws on its own mix of training data and live search.
The evidence matters as much as the answer. When ChatGPT searches, OpenAI says it may share the query with third-party search providers such as Bing, and the response shows inline citations and a list of sources (OpenAI Help Center: ChatGPT search). OpenAI also runs its own crawler, OAI-SearchBot, to surface websites in ChatGPT’s search features (OpenAI: overview of OpenAI crawlers). Google states that a page needs no special optimization to appear in AI Overviews or AI Mode, but it must be indexed and eligible to show with a snippet (Google Search Central: AI features and your website).
A full read therefore covers three layers: the answers themselves, the search results for the name and its variants, and the shared sources both draw on, such as review platforms, knowledge panels, company records and the press. At Buzz Dealer this read runs on Atom, our audit and reporting system. It scores every read per market and per language, and a specialist reviews each report before it is delivered.
Markets and languages change the answer too. The same question asked in German from Frankfurt and in English from London can bring back different sources and a different verdict, because the engines search local results and local press. For a brand that sells in several markets, the audit reads each market in its own language.
The three pillars: Sentiment, Discoverability and Authority
Sentiment
Sentiment is the tone and the wording the assistants use. Star ratings matter less here than sentences. “A well-reviewed platform with responsive support” and “a platform with a history of withdrawal complaints” can describe two brands with the same average rating. The audit records the words, the claims attached to them and whether those claims are current.
Discoverability
Discoverability is whether you are named when it counts. A brand can be described accurately when someone asks for it by name and still be absent when a buyer asks for the best option in its category. The audit tests both kinds of question and records your position against the names that appear instead.
Authority
Authority is the strength of the sources behind the answers. Assistants lean on pages they trust: established publications, review platforms, official records and well-maintained company pages. When the sources are thin, old or written by others, the answers follow them. The audit lists the domains cited for each question and marks the ones you can influence.
The same three pillars frame measurement beyond a single audit, as we set out in what brands should be measuring now.
Brand audit and individual audit
A brand audit reads a company: what the engines say and cite, how the search results for the name break down, the review platforms and the authority signals, all graded in the three pillars with a verdict for each.
An individual audit does the same read for a person, with the questions that matter for people. Do the engines recognize the right person, or confuse them with someone who shares the name? Are the facts right: role, company, history? How do results, news and AI answers portray them? For founders and executives the two audits often run together, because buyers ask about the company and the people behind it in the same session.
A first check you can run yourself
Before a full audit, a short manual check shows whether there is a problem worth measuring:
- Ask three assistants the same two questions: who or what is [name], and can it be trusted?
- Save the answers with the date and the assistant’s name.
- Open the sources each one cites and note which of them you would want a buyer to read.
- Note what is missing: facts that are out of date, achievements that never appear, competitors named where you are not.
This is the same first check described in our guide to the online reputation audit. It shows the shape of an issue. It cannot show the full picture across engines, markets and languages, or which source change would move which answer.
What the audit report gives you: a verdict per pillar and a prioritized list of gaps
The report turns a large number of answers into a small number of decisions. For each pillar it gives a verdict and the evidence behind it: the questions asked, the answers given and the sources cited.
The part that matters most is the gap list, in priority order. Each gap names what is wrong or missing, which answers it affects, which sources cause it and what would change it. That order is what turns an audit into a plan.
A good report also separates what you can correct from what you can only outweigh. A wrong fact on your own page is a quick correction. A critical article on an independent site stays where it is, and the work is to give the engines stronger and more current sources to read next to it.
See how you show up across AI and search
Enter your brand and website to generate an automated Brand Visibility & Reputation Audit. It measures how clearly your brand is understood, discovered, and trusted across search and AI, including sentiment, discoverability, and authority.
Get a free brand audit.
From audit to fixes
Fixes go to the sources. There is no form that changes what ChatGPT says about you: what changes the answer is the material it reads. The usual order:
- Wrong facts that cost deals. An outdated price, an old license status, a former executive still named as CEO. Correct the source pages first, starting with the ones cited most often.
- Missing answers to buyer questions. Where competitors are named and you are not, build the pages that answer those questions, on your own site and in independent sources.
- Sentiment. Tone moves last, as accurate and current sources come to outweigh old ones.
Illustration: a SaaS tool is described with last year’s pricing because the most cited comparison article predates its new plans, and its own pricing page never states the current plans in plain text. The fix is in those two pages.
Illustration: a fintech app is named with an old license status that a news piece reported two years ago. The fix is an up-to-date official record and coverage of the current status.
Illustration: a professional services firm is absent from category answers that name three rivals. The fix is independent coverage and listings that place the firm in that category.
Sentiment has its own mechanics, explained in how sentiment shapes AI recommendations. When the gaps call for sustained work across sources and markets, that is AI reputation management.
Audit, tracking and share of voice: how they fit
The audit is the baseline. It goes deep at one point in time, usually once a quarter, or before and after a major change.
Tracking is what follows: the same set of questions repeated on a schedule to see whether the answers move. Our guide to tracking your brand’s visibility in ChatGPT covers how to set it up.
Share of voice is one metric inside Discoverability: how often you are named against competitors across a set of questions. It is explained in AI share of voice.
Between audits, two free reports show how your own pages appear in AI answers. Google counts AI Overviews and AI Mode traffic in the Search Console Performance report, under the Web search type. Bing Webmaster Tools reports citations, cited pages and grounding queries in its AI Performance report.
AI visibility audit: FAQ
How do you measure your AI visibility?
Ask AI assistants a fixed set of questions about your brand and your category, record the answers and the sources they cite, and grade them. Grade three things: how the answers describe you (sentiment), whether you are named for the questions buyers ask (discoverability) and how strong the cited sources are (authority). Repeat the same questions over time to see what moves.
What is an AI visibility report?
An AI visibility report is the written result of an AI visibility audit. It shows what assistants such as ChatGPT, Gemini, Copilot, Perplexity and AI Overviews say about a brand or a person, which sources they cite, a verdict for sentiment, discoverability and authority, and a prioritized list of gaps with the sources that would change each answer.
What is AI visibility?
AI visibility is how present, and how accurately described, a brand or a person is in the answers AI assistants give. It covers whether you are named when people ask about your category, how you are described when they ask about you directly, and which sources the assistants rely on to say it. High visibility with wrong facts still costs deals.
How do I check the AI visibility of my brand?
Ask three assistants who your brand is and whether it can be trusted, then ask them for the best options in your category. Save the answers with the date, open the sources they cite and note what is wrong or missing. For a graded read across engines, markets and languages, request a full audit report.
How do I monitor brand visibility in AI?
Repeat the same prompts on a schedule, monthly for most brands, and record whether you are named, how you are described and which sources are cited. Add the Search Console Performance report and the Bing Webmaster Tools AI Performance report for your own pages. Run a full audit each quarter or after any major change.
How often should you run an AI visibility audit?
Once a quarter suits most brands, with monthly spot checks in between. Run one sooner after a major change: a rebrand, a funding round, a leadership change, a product launch, a crisis or entry into a new market. Each audit should repeat the questions of the last one, so the results can be compared.