Guide
AI visibility: what it is, how to measure it, how to improve it
People now ask ChatGPT, Perplexity, Gemini, Claude and Google’s AI Mode for recommendations. This guide explains how those answers choose brands, how to measure a brand’s place in them honestly, and what actually moves the numbers.
Last updated
What is AI visibility?
AI visibility is how often, and how well, AI assistants mention, recommend and cite a brand when people ask questions the brand should be part of.
When someone asks “who is the best emergency plumber in Austin?” or “which SEO tool tracks ChatGPT?”, an assistant writes an answer, often naming a handful of brands and linking to a few sources. Being in that answer, being described correctly, and being the source it links to are the three things AI visibility measures.
It matters because these answers increasingly come before a click: a buyer can form a shortlist without visiting a single website. A brand that is missing, or described with a wrong price or location, loses that buyer without ever seeing them in its analytics.
AI visibility, AEO, GEO and SEO: what’s the difference?
SEO is about ranking in search results; AEO and GEO are names for the work of appearing in AI-generated answers; AI visibility is the outcome you measure. The work overlaps heavily.
- SEO (search engine optimization): making pages rank in search results and earn clicks.
- AEO (answer engine optimization) and GEO (generative engine optimization): making a brand appear, be recommended and be cited in answers written by AI assistants. The terms are used interchangeably.
- AI visibility: the measurement: mention, recommendation and citation rates across the questions that matter, over time.
The overlap is large. Google states that its AI Overviews and AI Mode use the same crawling and indexing as Search, and assistants with web search retrieve pages from search indexes before they write. A page that isn’t crawlable, indexed and trustworthy is rarely cited. What is new is the measurement, and the importance of being mentioned on other sites that assistants read.
Why can’t one screenshot measure AI visibility?
Because AI answers vary from one ask to the next. One answer is an anecdote; a rate over many sampled answers, with its likely range, is a measurement.
Ask the same question three times and you may get three different lists of brands. Assistants generate each answer fresh, and assistants with web search may retrieve different sources each time. A screenshot shows what happened once; it says nothing about how often it happens.
The useful question is “in what share of answers is the brand mentioned?”. Ask the same set of questions several times, on a schedule, in the same market, and count. If the brand is named in 19 of 50 answers, the mention rate is 38%.
How do you measure AI visibility properly?
Track a fixed set of buyer questions, sample each one several times per engine on a schedule, and report each rate with a confidence range and the method used.
The six metrics that matter
- Mention rate
- The share of sampled answers that name the brand at all. The broadest signal, and the first to move.
- Recommendation rate
- The share of answers that explicitly recommend the brand, not just list it. Judged against a fixed rubric.
- Citation rate
- The share of answers that link to one of the brand’s own pages as a source. Citations send visitors; mentions alone often don’t.
- Share of voice
- The brand’s mentions as a share of all mentions of the tracked brands (the client and its competitors) for the same questions.
- Sentiment
- How the brand is framed when it is mentioned, from −1 (warned against) to +1 (recommended as the best fit).
- Accuracy
- For questions about the brand itself: the share of stated facts (prices, locations, services) that are correct.
Always show the range
A rate from a sample is an estimate. With 50 answers, a 38% mention rate is likely somewhere between about 26% and 52%, so the next measurement could land anywhere in that range by chance alone. The Wilson score interval is a standard way to compute the likely range for a proportion, and it behaves well with small samples. When two periods’ ranges overlap, the change may be noise; when they don’t, something has probably moved.
Label the method
Sampling through an engine’s API is not the same as what every user sees: personalisation, location and signed-in history differ. That is fine as long as the method is the same every time and labelled, so a change in the number means a change in the answers, not in how they were collected.
One headline number, four parts
A single score is convenient for a report, but it can hide which part is broken. In ZORO SEO the AI Visibility Score is 40% mention rate, 30% citation rate, 20% share of voice and 10% sentiment, always shown next to its four components.
Which questions should you track?
The questions real buyers ask, from facts about the brand itself to the broadest category questions, grouped so you can start where you can win.
Good sources of questions are the business’s own services and locations, People-also-ask questions in Google, and the grounding queries Bing reports for Copilot. A useful way to organise them is a ladder from easiest to hardest:
| Level | Example | Aim |
|---|---|---|
| L1 · Brand truth | “What does Acme Plumbing charge for a call-out?” | Correct facts in every answer about the brand itself. |
| L2 · Niche long-tail | “Plumber in Austin who fixes tankless water heaters on Sundays” | Be named for very specific needs, where competition is thin. |
| L3 · Alternatives and comparisons | “Acme Plumbing vs Roto-Rooter” | Appear, and be described fairly, when buyers compare options. |
| L4 · Segment best-of | “Best emergency plumber in Austin” | Be recommended for a segment or a location. |
| L5 · Category head | “Best plumbing company” | A share of voice in the broadest, most contested questions. |
Start with brand truth and niche questions, where a smaller brand can realistically be named, and move up the ladder as those goals are met.
How do AI assistants choose which brands and sources to include?
Assistants with web search retrieve pages for the question, then write an answer from what they read. Brands that are clearly described on their own site and mentioned on sources the assistant trusts are more likely to be included.
Each engine retrieves differently, and they often cite different pages for the same question, so it pays to look at each engine separately. Common patterns are consistent, though:
- Retrieval comes first. A page has to be crawlable and indexed to be read. Google says its AI features rely on its normal index; other assistants rely on their own or partners’ indexes, which is why Bing indexing and AI crawlers’ access in robots.txt matter.
- Specific, verifiable content gets used. In the GEO study by researchers at Princeton and others (KDD 2024), adding statistics, quotations and cited sources to pages increased how visible they were in generative answers, while keyword stuffing did not help.
- Other sites’ opinions count. Review sites, comparison articles, directories and community discussions are frequently cited for “best” and “alternatives” questions. A brand that appears there, with genuine reviews, is easier to recommend.
- Consistency helps. The same name, services, prices and locations everywhere make a brand easier to describe correctly.
What improves AI visibility?
The same things that earn trust from people: correct facts, clear pages that answer real questions, and genuine mentions on the sites assistants read.
- Fix the facts. Put correct prices, services, locations and hours on the site, consistently, and correct them where assistants get them wrong.
- Let crawlers in. Allow search and AI crawlers in robots.txt, make sure pages render text without JavaScript, and submit a sitemap to Google and Bing.
- Answer the questions buyers ask. One page per real question, with the direct answer first, then the detail, figures and sources.
- Publish honest comparisons. Pages that compare options fairly, including when another option fits better, are useful to buyers and to assistants.
- Earn mentions where assistants look. Complete profiles on the review sites and directories that appear in answers for your topics, and ask real customers for reviews.
- Measure, then repeat. Track the same questions, compare ranges, and keep what moves the numbers.
What doesn’t work, or backfires?
Shortcuts: fake reviews, paid or invented mentions, keyword stuffing and AI-specific tricks. They break platform rules, and the research suggests they don’t help.
- Fake reviews and astroturfing break the rules of review sites and communities, and are removed. ZORO SEO won’t create tasks for them.
- Keyword stuffing performed worse than doing nothing in the GEO study.
- Special files or markup “for AI”. Google says none are required for its AI features. Files like llms.txt are harmless but not a shortcut.
- Judging by screenshots. Acting on one answer leads to chasing noise. Decide on rates and ranges.
How does ZORO SEO measure and improve AI visibility?
It researches the questions buyers ask, samples them on five AI engines on a schedule, reports every rate with its range, and turns the gaps into clear tasks for your team.
- Questions come from the client’s approved business profile, People-also-ask and Bing grounding queries, placed on the five-level ladder above.
- Each question is asked several times per engine per cycle (3 to 5, depending on the plan) on ChatGPT search, Perplexity, Gemini with Google Search, Claude with web search and Google AI Mode.
- Every rate is shown with its Wilson 95% range and the method version, and is labelled as API-sampled.
- Answers are checked against the client’s approved facts; a wrong fact becomes an alert and a task.
- The sites AI answers cite for your topics become tasks with clear titles, such as “Create the G2 profile” then “Get 10 genuine reviews on G2”.
See the walkthrough for the screens, or the blog for more articles.
Frequently asked questions
What is AI visibility?
AI visibility is how often, and how well, AI assistants such as ChatGPT, Perplexity, Gemini, Claude and Google AI Mode mention, recommend and cite a brand when people ask them questions the brand should be part of.
Is AI visibility the same as AEO or GEO?
They describe the same goal. AEO (answer engine optimization) and GEO (generative engine optimization) are names for the work of improving how a brand appears in AI answers; AI visibility is the outcome you measure.
Does SEO still matter if people use AI assistants?
Yes. Google says its AI features draw on the same crawl and index as Search, and assistants with web search retrieve pages from search indexes. Pages that can’t be crawled, indexed or trusted are rarely cited.
Why do AI answers change every time I ask?
Assistants generate each answer fresh and may retrieve different sources each time, so the same question can name different brands. That is why a single answer or screenshot is not a measurement; you need repeated samples and a rate with a range.
How many samples do you need to trust an AI visibility rate?
More samples give a narrower range. With 50 answers, a 38% mention rate is likely somewhere between about 26% and 52%. Tracking the same questions on a schedule and comparing ranges, not single numbers, avoids over-reading noise.
Do I need an llms.txt file or special markup to appear in Google’s AI features?
Google’s guidance says no special files or markup are required for AI Overviews and AI Mode; the usual SEO fundamentals apply. An llms.txt file is harmless and some tools read it, but it is not a ranking factor for Google.
Can you guarantee a brand will be recommended by ChatGPT?
No. Nobody controls what an assistant says. You can make the brand easier to find, understand and trust, and measure whether that moves the numbers.
How long does it take to improve AI visibility?
Correcting wrong facts and fixing crawler access can show up within weeks, as assistants re-read pages. Being recommended for competitive questions usually takes months, because it depends on other sites mentioning and reviewing the brand.
Which AI engines does ZORO SEO track?
ChatGPT search, Perplexity, Gemini with Google Search, Claude with web search and Google AI Mode. You can also upload Bing’s AI Performance report and Search Console’s Generative AI report.
Sources
- Google Search Central: AI features and your website: how AI Overviews and AI Mode relate to Search, and what site owners need to do
- Aggarwal et al., “GEO: Generative Engine Optimization” (KDD 2024): statistics, quotations and cited sources increased visibility in generative answers
- Wilson score interval: the confidence range used for every rate
Measure AI visibility for your clients
Five engines, confidence ranges, and the tasks that move the numbers.