Resources · last reviewed 2 November 2026 · by Wahid Ryland
How does AI search visibility tracking work?
AI search visibility tracking asks AI engines the questions your buyers ask, records every answer, and counts two things: how often the engine names your company, and how often it cites your website as a source. Because AI answers change from run to run, a tracker asks each question several times and reports rates, not single results.
What gets measured
- Named
- Your company’s name appears in the answer text. The model produced it, with or without a source.
- Cited
- Your own domain appears in the engine’s list of sources. The engine read your website to write that answer.
These are different facts. A company can be named often and cited rarely, which means the engines know the brand but are not reading its site. Get AI Cited reports the two separately on every page.
How a measurement runs
- 01
11 buyer questions
Written the way buyers ask; no measured company named
- 02
6 AI engines
ChatGPT, Perplexity, Claude, Gemini: 3 runs each. Grok and Google AI Overviews: 1 run
- 03
154 answers kept
Full text and source list for every run
- 04
Each answer checked
Named: your name in the text. Cited: your domain in the sources
- 05
Cited Score
70% cited, 30% named, six engines averaged equally
The score is published only when ChatGPT, Perplexity, Claude and Gemini each answered at least 60% of their runs. A missing engine would make a company look invisible when the engine simply failed, so in that case no score is published.
- A fixed set of buyer questions. Each category has 11 questions: six recommendation, three comparison, two category. No question names a company being measured, because naming one guarantees it a mention.
- Six engines. ChatGPT, Perplexity, Claude, Gemini and Grok are asked through their official APIs with web search on. Google AI Overviews is read from a live US Google results page.
- Repeated runs. Each question runs three times on the four chat engines and once on Grok and Google AI Overviews: 154 answers per scan.
- Every answer kept. The full text, the source list, the companies named and the run number are stored, so any figure traces back to the answers behind it.
- Matching by hostname and whole words. A source counts as cited only if its host is your domain or a subdomain of it, and a name only as a whole word.
Why one engine is not enough
Engines do not share their sources. In our measurements across two B2B categories, ChatGPT shared only 12–29% of its cited sources with each other engine (the research). A company can be cited by Perplexity and absent from ChatGPT on the same question, so a tracker that checks one engine misses most of the picture.
Two things that surprise people
- Some answers come from memory. Engines sometimes answer without searching. On 5 of 11 payments questions, Claude answered from memory in every run (the research). Those answers still name companies, but nothing in them can be cited.
- Google does not always show an AI Overview. A question with no Overview is recorded as a result, and AI Overviews is scored only on the questions where one appeared.
What to look for in a tracker
- It reports cited and named separately.
- It asks each question more than once, and says how many times.
- It names the engines and model versions it uses.
- It keeps the answers, so you can read what the engine actually said.
- It publishes its method, including when it withholds a result.
Agencies comparing tools can start with AI visibility tracking tools for agencies, compared.
Questions
How often should AI visibility be measured?
Weekly is enough for most companies. Engine answers move from day to day, so a single reading is noisy; the trend across scans is what matters.
Can I track AI visibility by asking ChatGPT myself?
You can see one answer. A tracker asks the same questions repeatedly across six engines, keeps every answer and counts the results, which a manual check cannot do consistently.
What is the difference between being cited and being named?
Named means the engine wrote your company's name. Cited means it used your website as a source for the answer.
Does Get AI Cited track Google AI Overviews?
Yes. Since 31 October 2026, Google AI Overviews and Grok are part of every Cited Score, alongside ChatGPT, Perplexity, Claude and Gemini.
Where is the full method?
On the methodology page at getaicited.co/en/methodology, including the weights, the engines' model versions and the dated list of changes.
The full method, with weights, model versions and every change dated: Cited Score methodology.