Methodology
How the AI Citation Score is measured
The score is measured, not modelled: every number comes from a real answer returned by an AI engine and the sources it cited.
Where the data comes from
- ChatGPT: OpenAI Responses API with web search turned on. We read the URLs ChatGPT attaches to its answer as citations.
- Perplexity: Perplexity's API. We read the search results Perplexity returns as the sources of its answer.
- Gemini: Gemini API with Google Search grounding. We read the web sources listed in the answer's grounding metadata.
- Google AI Overviews: A live Google results page fetched through DataForSEO. We read the references listed in the AI Overview block.
Questions about the score
What is the AI Citation Score?
The percentage of AI answers that cite your website as a source. If 4 engines answer 10 prompts (40 answers) and 10 of those answers link to your domain, your score is 25%.
Does an AI model decide whether I was cited?
No. We never ask a model to judge its own answer. A prompt counts as cited only when one of the engine's returned source URLs is on your domain or one of its subdomains.
What happens when an engine fails or times out?
That answer is recorded as an error and left out of the score entirely. Errors are never counted as "not cited", so an outage can't lower your score.
How are competitors chosen?
Competitors are the other domains cited in the same answers, counted at most once per answer and ranked by how many answers cite them. Your own domain and subdomains are excluded, and so are search engines, maps, social networks, YouTube and Wikipedia, because those are sources rather than businesses you compete with.
How often is it measured?
Tracked prompts are re-run on every engine each week. Each report uses the latest successful answer for every prompt and engine, and the weekly change compares that with the latest answers from 7 days earlier.
How is the free checker different?
The free AI Citation Score checker is a quick snapshot: it checks Google AI Overviews for up to 3 keywords and caches the result for 7 days. Paid plans track your own prompts across all four engines every week.
Why can my score change between weeks when I changed nothing?
AI answers are not fixed. The engines re-search the web, and different sources rotate in and out. That is why we track the same prompts over time instead of trusting a single run.