How to Track AI Visibility: Cited, Named or Missing
- AI Visibility
- Measurement
- GEO
To track AI visibility, pick 20 to 50 real buyer questions, run each one through ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews on a fixed schedule, and record whether your site is cited as a source, named in the text, or missing. Weekly checks with the same wording and settings turn noisy single answers into a trend. Then fix the questions where you are missing or only named, starting with those closest to a purchase.
On this page
What is AI visibility?
AI visibility is how often AI answer engines bring up your site or your brand when people ask the questions you want to win. It is measured one question at a time: does the answer cite your page as a source, name your brand in the text, or leave you out?
You have to track it on purpose, because your usual reports barely see it. Google Analytics only records the visits that happen, and many AI answers end without a click. Search Console folds Google's AI features into its regular numbers, and the other engines never appear there. A site can look healthy in both while a competitor is recommended in every answer.
AI visibility is how often AI answer engines cite your site or name your brand for the questions you want to win, measured question by question across ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews.
What is an AI monitoring system?
In machine learning, an AI monitoring system watches a model in production for errors and drift. In search marketing it means the reverse: a system that watches what AI models say about your brand. The second kind is the one you need to track AI visibility.
It has four parts, and a spreadsheet is enough to start. Tools automate the running and logging, but the method is the same.
- A fixed list of questions, worded the way buyers ask them.
- A fixed set of engines: ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews.
- A schedule, so each check can be compared with the last one.
- A log of the date, engine, question, outcome and who was cited instead.
Which questions should you track?
Track the questions people ask before they know your name. Brand questions, such as what is [your brand], mostly tell you whether the engines know you exist. Category, problem and comparison questions are where an engine decides which brands to recommend, so they deserve most of the list.
Start with 20 to 50 questions from real sources: your Search Console queries, the People Also Ask boxes for your main keywords, and the words customers use on sales calls. Write each as a full sentence, the way people type into ChatGPT, and leave your brand name out so the test shows whether the engine reaches for you unprompted.
Tag each question by type and by how close it sits to a purchase.
- Category: best [product type] for [use case].
- Problem: how do I [the job your product does].
- Comparison: [competitor] alternatives, or [option A] vs [option B].
- Brand, kept in its own group: what is [your brand], is [your brand] worth it.
Track the category, problem and comparison questions buyers ask before they know your brand, written as full sentences without your brand name, because those are the answers where AI engines choose who to recommend.
How do you check ChatGPT, Perplexity, Gemini, Claude and AI Overviews?
Run every question in a clean session so past chats do not shape the answer. Keep the wording, location and mode the same each time, and record the date, because the engines change week to week. Each engine shows sources differently, which decides whether a result can count as cited.
Google AI Overviews cannot be prompted: you search the question on Google and see whether an overview appears. Many searches do not show one, so log that as its own outcome rather than as missing. Google AI Mode, the chat-style mode inside Google Search, is worth checking the same way for your top questions.
- ChatGPT: use a temporary chat so memory stays out of it, with web search on so its sources are shown.
- Perplexity: it searches the web for almost every answer and numbers its sources.
- Gemini: it shows sources on some answers and not on others, so expect more named results here.
- Claude: turn web search on, because without it Claude answers from training data and can name you but cannot cite a page.
- Google AI Overviews: search in a private window from the same location each time, then check the links inside the overview.
Cited, named or missing: how do you score each answer?
Give every answer one of three outcomes. Cited means your site is one of the sources the answer links to. Named means your brand appears in the answer text, but none of the sources is yours. Missing means you are not in the answer at all. Also note who was cited instead and, when you win, which of your URLs was cited.
The split between cited and named matters because each needs a different fix. Named means the engine knows your brand but trusted someone else's page for the facts. Cited means your own page shaped the answer and gives the reader a direct path to you.
Roll the outcomes into two numbers: cited rate, the share of answers that cite you, and mention rate, the share that cite or name you. Split both by engine and by question type. Tools that report an AI visibility score each use their own formula, so check how it is built before comparing tools.
Score every AI answer as cited (your site is a linked source), named (your brand appears with no link to you) or missing, because each outcome needs a different fix.
How often should you track AI visibility?
Weekly suits most sites. AI answers vary from run to run, so one check is a snapshot that can mislead either way, while the same questions checked weekly for a month or more show a real trend. If you check by hand, monthly is a realistic minimum.
Read the pattern, not the single answer. A question where you are cited three weeks out of four is a win to protect; one where you appeared once in two months is not a win yet. After you change a page, give it several weekly checks before you judge the fix.
Fifty questions across five engines is 250 answers every week, a lot to check and log by hand. SmashSERP runs that loop on a schedule: each week it checks every tracked question in ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews and records cited, named or missing, while its Rank Tracker logs your Google positions on the same weekly cycle. Details are at /ai-visibility-tool.
Track AI visibility weekly with the same questions and settings, because AI answers vary from run to run and only repeated checks over several weeks show a real trend.
How do you track Google AI Mode traffic?
You cannot pull it out as a clean number. Google has reported AI Mode clicks and impressions inside the Search Console Performance report, blended with regular web search, and AI Overviews are counted the same way. AI Mode runs on google.com, so its visits generally reach Google Analytics as ordinary google / organic sessions.
So track AI Mode from the answer side. Run your top questions in AI Mode on the same schedule as the other engines and log cited, named or missing. Then use Search Console as a warning light: when a page's impressions climb while its clicks stay flat, check its main questions in AI Mode and AI Overviews by hand.
How do you see AI referral traffic in GA4?
Clicks from the chat engines are easier to isolate. ChatGPT often adds utm_source=chatgpt.com to the links in its answers, and visits from Perplexity, Gemini and Claude usually carry their own domain as the referrer. A custom channel group in GA4 gathers them in one place.
Treat the result as a floor, not a full count. Many AI answers end without a click, and visits from some apps arrive with no referrer and land in Direct. Referral data shows which pages AI engines send people to; your answer log shows how often you are in the answer at all.
- In GA4, go to Admin, then Data display, then Channel groups, and create a new group.
- Add a channel named AI assistants, set its rule to Source matches regex, and use chatgpt\.com|perplexity\.ai|gemini\.google\.com|claude\.ai as the value.
- Drag the new channel above Referral, because GA4 assigns each session to the first channel it matches.
- Compare the channel's landing pages with your answer log to see which cited pages earn visits.
What should you do with the results?
Sort the log by outcome and work the questions closest to a purchase first; each outcome points to a different job, listed below. The engine-specific playbooks are in /blog/how-to-get-quoted-by-chatgpt and /blog/how-to-optimize-for-ai-overviews.
Keep the log going after each fix. When a cited page slips back to named or missing, check what changed: a competitor may have published a sharper answer, or your page may be out of date. Once a month, roll the log into a short AI visibility report: cited rate by engine, questions gained and lost, and the pages you changed.
- Missing: write or expand a page that answers the question in its first sentence, under a heading that matches it.
- Named, not cited: rewrite the opening of your page so it answers more directly than the cited source.
- A competitor cited every week: study their cited passage, match its clarity and add what it leaves out.
- Cited: keep the page current and do not bury the answer in a redesign.
- No AI Overview shown: keep tracking, but put classic ranking work for that query first.