AI visibility is how often, and in what way, AI assistants such as ChatGPT, Gemini, Perplexity and Google AI Overviews name your brand when a buyer asks a question about your category. A brand can be mentioned, cited as a source or recommended. It can also be absent or described wrongly. You measure it as a share of answers across repeated runs, not as a ranking position.
Key takeaways
- AI visibility has five states: absent, misdescribed, mentioned, cited and recommended. Track them separately.
- Buyers already ask AI: 82% of B2B software buyers in a G2 survey sourced software recommendations from an AI chatbot in the last 24 months (G2, July 22, 2026).
- AI has not replaced search: AI tools made up 1.65% of US desktop browsing events in Q1 2026 (Datos, via PPC Land, April 27, 2026).
- An AI answer has no fixed positions. Measure mention rate: the share of answers naming your brand over many runs.
Mentioned, cited or recommended: what is the difference?
These words are often used as synonyms, but they are separate states. The table uses an illustrative example, not real data: an invented invoicing tool called Northbill and the prompt "What is the best invoicing software for freelancers?"
| State | What it means | Illustrative answer | What to do |
|---|---|---|---|
| Absent | The answer names other brands, not yours | "Popular options include Brand A and Brand B." | Record who is named instead |
| Misdescribed | The brand is named, but a fact is wrong | "Northbill has no free plan." (It has one.) | Find the likely source page and correct it |
| Mentioned | The brand appears, with no advice to choose it | "Other tools include Northbill." | Count it, but not as a win |
| Cited | The assistant links to or names your page as a source | A link to a Northbill post under an answer recommending Brand A | Note which pages earn citations |
| Recommended | The assistant advises choosing the brand | "For freelancers, Northbill is a strong choice." | Track this share separately |
A citation without a recommendation is possible: your article can be the source for an answer that advises a competitor. So is a mention without any link to your site.
Why does AI visibility matter now?
Because fewer searches end in a click, and buyers report taking recommendations straight from AI answers.
- Most Google searches end without a click. In January to April 2026, 68.01% of US Google searches ended without a click, according to SparkToro's analysis of Similarweb panel data, published June 8, 2026. The panel covers desktop and mobile browsers, not the Google app.
- Results pages with AI summaries get fewer clicks. In Pew Research Center's analysis of 900 US adults' browsing in March 2025, published July 22, 2025, users clicked a traditional result in 8% of visits to a Google results page with an AI summary, against 15% without one. This is an observed difference, not an experiment.
- Usage is large. Google said on May 19, 2026 that AI Mode had passed one billion monthly users a year after its debut.
- Buyers ask AI for shortlists. In G2's survey of more than 1,000 B2B software buyers and decision-makers, published July 22, 2026, 82% had sourced software recommendations from an AI chatbot in the last 24 months. As a source shaping the vendor shortlist, AI chatbots (37%) came just behind review sites (38%). G2 is itself a review site, and the release does not say where or when the survey ran.
How big is AI search really?
Smaller than the headlines suggest.
AI tools accounted for 1.65% of desktop browsing events in the US in Q1 2026, up from 1.31% a year earlier, and for 1.72% in the EU and UK. The figures are from the Datos State of Search report, as reported by PPC Land on April 27, 2026; the full report is behind a sign-up form. The data is desktop only. The write-up also quotes lower monthly shares (0.93% for the US in March 2026) without explaining the gap; either way, the share is below 2%.
A second check from the SparkToro analysis above: only 0.34% of US Google searches went to AI Mode in January to April 2026. That does not contradict Google's billion users: one figure counts people, the other searches in US browsers.
So "AI has replaced search" is false. But "AI sends almost no traffic, so it does not matter" does not follow either. One vendor study suggests that visits after an AI mention rarely carry an AI label.
Profound, which sells AI visibility software, published the study on July 1, 2026, using more than 2 million AI conversations from a US opt-in panel, January to June 2026. It took cases where an assistant named a brand the user had not typed, then checked whether that user visited the brand's site within seven days.
For ChatGPT, 6.39% of such mentions were followed by a visit, against a baseline of 4.33% forecast from the same users' earlier behavior. In June 2026 (a partial month), 2.47% of those visits carried an AI-referral tag. More than 97% did not.
This is one vendor's panel, a site-visit study rather than a sales study, and not a randomized experiment. Many untagged visits would have happened anyway, which is what the baseline measures. The study suggests that referral reports undercount visits that follow AI answers; it gives no precise multiplier.
How is AI visibility different from search rankings?
A search results page has positions. An AI answer does not.
In research by SparkToro and Gumshoe, published January 27, 2026, 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI answers (AI Overviews, or AI Mode when no overview appeared) a combined 2,961 times. For ChatGPT and Google, the chance that any two of 100 responses gave the same list of brands was below 1 in 100; Claude repeated a list slightly more often. The same list in the same order was rarer still: about 1 in 1,000.
SparkToro's Rand Fishkin concluded that a "ranking position in AI" is not a meaningful number, while the percentage of answers that include a brand, measured across dozens to hundreds of prompts run multiple times, is a reasonable metric. Gumshoe is an AI tracking vendor, as Fishkin himself discloses.
One screenshot of one answer therefore proves nothing. Why AI answers change every time explains how many runs you need.
Which AI visibility metrics should you track?
| Metric | Definition | Caveat |
|---|---|---|
| Mention rate | Share of answers that name your brand over repeated runs | Needs enough runs; report a range, not a single number |
| Recommendation rate | Share of answers that advise choosing your brand | Always at or below the mention rate |
| Share of voice | Your mentions as a share of all brand mentions in the same answers | Depends on the competitors and prompts you include |
| Cited sources | Pages and domains the answers link to or name | A cited page is not a recommended brand |
| Sentiment | Whether the brand is described positively, neutrally or negatively | Automated labels can be wrong; read a sample |
For the formula and its pitfalls, see AI share of voice.
How do you start measuring AI visibility?
- Check by hand. Open three or four assistants in a clean session and ask buyer questions that do not contain your brand name. Note the state of each answer. How to check what ChatGPT and Gemini say about your brand gives the full procedure.
- Fix a prompt set. Write the questions down and keep them unchanged, so later results stay comparable.
- Repeat on a schedule. Run every prompt several times on each assistant. Compare shares over time, not single answers.
Done by hand, this stops scaling quickly. Brandometer automates it. It sends each prompt to 8 AI engines repeatedly, in clean sessions without personalization. It reports mention rate with confidence intervals, whether the brand is recommended or merely mentioned, sentiment, the competitors named instead, cited sources and change over time.
It also lists suggested actions. Treat them as hypotheses to test: no tool can promise to change what an AI says.
Check your own brand for free
Run a free one-off check on the home page to see how AI assistants answer a question about your category.
New accounts get 200 bonus credits when the e-mail address is confirmed. There is no subscription: you top up credits when you want to measure, and they never expire.
FAQ
Is AI visibility the same as GEO?
No. AI visibility is the outcome: how assistants present your brand. GEO (generative engine optimization) and AEO (answer engine optimization) are names for the work aimed at changing it, and much of that work overlaps with ordinary SEO and PR. GEO vs SEO vs AEO compares the terms.
Can I see AI visibility in Google Search Console?
Only a small part of it. Search Console's generative AI performance reports, announced by Google on June 3, 2026, show how often URLs from your site appeared in AI Overviews, AI Mode and generative AI features in Discover. Google says they reached all sites worldwide on August 31, 2026. They had no click data at that point, Search Engine Journal reported the same day. An impression count for your pages does not tell you whether the answer named or recommended your brand, and the reports cover Google only.
How often should I check AI visibility?
Often enough to see a trend, and always after a change. A single check is a snapshot, because answers vary from run to run. Measure a baseline, repeat it on a fixed schedule such as monthly, and measure again after you update key pages.