Skip to content
GET-GEO.AI
/
← All guides

// guide

How do you measure AI visibility and citations?

Dmitry Filippov

Written by: Dmitry Filippov, Founder, GET-GEO.AI
Published: 2026-07-31 · Updated: 2026-11-18

// short answer

Measure AI visibility with four inputs: citation rate across a fixed prompt set sampled repeatedly, share of voice against competitors on those same prompts, AI-crawler hits in server logs as an early check, and referral traffic from assistant domains as the commercial outcome. Single checks are unreliable because answers vary between runs.

  • AI visibility is measured as frequency, not position: out of N runs of a fixed prompt, how often you were named, tracked through four metrics from citation rate to referrals.
  • Start with a fixed set of twenty to fifty commercially relevant prompts, split into four types: category, comparison, branded and problem-led questions.
  • Plan three to five clean-session runs per prompt with frozen wording: in our lawyer study of 22 September 2026 (72 answers), two ChatGPT runs of one prompt shared on average 3.9 of 5.6 firm names.
  • Start log checks with four AI crawlers (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot): a fetch shows the page was reached, not that it will be cited.
  • Assistant referrals are a lower bound: on get-geo.ai in the week of 30 September–7 October 2026, Vercel's referrer report showed roughly a tenth as many ChatGPT visitors as the utm_source=chatgpt.com tag did.

Why rank tracking does not translate

A ranked list is stable enough that checking position once a day means something. A generated answer is produced fresh each time and varies between runs and with the form of the question, so checking once tells you what happened once.

In our lawyer study of 22 September 2026 (12 prompts for Houston, Miami and London, two runs each in ChatGPT, Perplexity and Google AI Mode, 72 answers), two runs of the same prompt kept on average 3.9 firm names of about 5.6 in ChatGPT, 1.7 of 3.7 in Perplexity and 1.4 of 4.2 in AI Mode, counting spelling variants as one firm. Perplexity's average includes three prompts where it named no firm. This is one niche on one evening, and the spread in your category may differ.

The unit of measurement has to change from position to frequency: out of N samples of a given prompt, how often were you named.

Build the prompt set first

Everything else depends on having a fixed list of questions that matter commercially. We advise starting with twenty to fifty, drawn from what buyers actually ask rather than from keyword tools.

Keep question types separate: category questions where you want to appear among the options; comparisons that already name a competitor; branded questions that test how you are described; and questions about a problem your content can answer.

  • Category: 'best GEO agency for B2B SaaS'
  • Comparison: 'X vs Y for AI visibility'
  • Branded: 'what does GET-GEO.AI do'
  • Problem-led: 'why is my brand not cited by ChatGPT'

The four metrics worth tracking

Citation rate is the share of prompts in a fixed set where an assistant links to one of your pages, counting a prompt if the link appears in at least one of its runs; a brand mention without a link counts towards share of voice instead. Share of voice is the same measurement expressed against the competitors appearing alongside you, which is what tells you whether a flat number is good or bad.

Crawler activity in server logs is the leading indicator: GPTBot, OAI-SearchBot, PerplexityBot and ClaudeBot hits confirm new content has been fetched. A fetch shows the page was reached, not that it will be cited, and we have no log series of our own that dates the gap between the two. What we have dated is publication to citation: a guide we published on 14 September 2026 was cited the next day by Perplexity (signed in) and by Google AI Mode, in AI Mode in both runs, on the one prompt it answers directly. One guide and one prompt show that the gap can be short; they do not set a norm. Referral traffic from assistant domains is the lagging indicator and the one that connects to revenue.

Check the limits of referral data

Assistant referrals are undercounted, and by how much depends on which report you read. On get-geo.ai in the week of 30 September–7 October 2026, almost one in five visitors counted by Vercel Web Analytics arrived with the utm_source=chatgpt.com tag, while Vercel's referrer report listed chatgpt.com for roughly a tenth as many visitors. That is one week on one site, and ChatGPT only; for other assistants, compare the referrer view with the tag view in your own counter. A user who reads an answer and later types your domain registers as direct traffic with no attribution at all.

Treat recorded referrals as a lower bound and watch the trend. Our guide to AI traffic in GA4 shows where these visits sit and why counters disagree: in that same week GA4 reported a little over a third of the visitors Vercel Web Analytics counted. We also advise watching searches for your brand in Search Console next to the AI figures, as an indirect signal rather than proof.

Repeat tests under consistent conditions

Run each prompt several times rather than once, in a clean session, logged out or with memory off, so that account history stays out of the result, and note the country you tested from. Record which sources were cited rather than only whether you appeared: the list of competitors and sources shows who holds the answer you want.

Keep the prompt wording frozen between rounds, because the form of the question can change the answer. In our local-business check of 27 September 2026 (ChatGPT and Google AI Mode, dentists and plumbers in Houston, Miami and London, 12 answers per question form from each assistant), "Can you recommend…" questions produced a ChatGPT map in 12 of 12 answers and 5.4 businesses named on average, while "How do I choose…" questions produced a map in 7 of 12 and 2.8 names. AI Mode showed a map in 11 of 12 answers to the first form and in none to the second. That is two niches, in English, on one day, and it compares two question types rather than small rewordings. Change the wording between rounds and you can no longer compare periods.

What each metric tells you
MetricSourceSignal typeReliability
Citation rateSampled prompt runsDirect outcomeGood, if sampled enough
Share of voiceSame runs, competitor setCompetitive positionGood
AI crawler hitsServer logsLeading indicatorHigh — server logs provide a record of requests
Assistant referralsAnalyticsLagging, commercialUndercounted; use the trend
Branded search volumeSearch ConsoleIndirect demandNoisy but useful
What each metric tells you

Related questions

How many samples per prompt are enough?

Enough that one unusual answer does not move the number. Start with three to five runs per prompt per round and add runs where answers vary most. In our lawyer study of 22 September 2026 (72 answers, two runs per prompt), Google AI Mode repeated about a third of its firm names between runs and ChatGPT about seven in ten.

Can I see AI citations in Google Search Console?

Partly. Search Console's Generative AI performance report, which Google rolled out to all websites by 31 August 2026, counts impressions of your links in AI Overviews and AI Mode by page, country, date and device. Clicks stay in the overall Performance report, and there is no query dimension, so which questions trigger a citation must be observed directly.

Do I need a paid AI visibility tool?

Not to begin with. A spreadsheet, a fixed prompt list and a disciplined weekly routine produce a defensible baseline. Tools mainly buy scale and scheduling once the prompt set grows; before paying, check that the tool keeps your prompt wording fixed and shows the sources behind each answer.

What does a good citation rate look like?

There is no universal benchmark: it depends on category competitiveness and how established the alternatives are. The meaningful comparison is your own trend over time, on the same prompts and engines, and your share against the specific competitors appearing in the same answers.

Related guides

Sources

  1. 01OpenAI — bots and crawler documentation
  2. 02Google Search Central — AI features and your website
  3. 03Search Console Help — Generative AI performance report (Search)
  4. 04Aggarwal et al., GEO: Generative Engine Optimization (2023), arXiv:2311.09735
  5. 05Our research: which lawyers ChatGPT, Perplexity and Google AI Mode recommend — 72 answers, 22 September 2026
  6. 06Our research: ChatGPT's map card in local niches, including the question-form check — 120 answers, 27 September 2026
  7. 07Our guide: how to see ChatGPT and Perplexity traffic in GA4 — our analytics for the week of 30 September–7 October 2026
  8. 08Our guide: how long does GEO take to show results? — dated observations of 14–15 September 2026

// share

LinkedInXReddit

How to cite this page

You may quote and reuse this content with attribution and a link to this page.

“How do you measure AI visibility and citations?” — GET-GEO.AI, 2026-11-18. https://get-geo.ai/en/guides/measure-ai-visibility