Methodology

How AI visibility is measured

AI systems can produce different responses to identical questions. Every number in Citeloop is therefore based on repeated observations across selected platforms and should be read directionally rather than as a deterministic search ranking.

01Repeated sampling

Each tracked question is asked again on a schedule across the supported assistants. Nothing is inferred from a single response, because a single response is not reproducible.

02Verbatim extraction

Brand names, competitors and cited domains are pulled from the answer text itself, then checked back against that text. Anything the extractor produces that does not literally appear is dropped.

03Entity resolution

Aliases, domains and location claims are resolved so a same-name rival in another town is not counted as you — and conflations are surfaced rather than silently merged.

04Frequency reporting

Results are expressed as “appeared in 7 of the last 10 observed answers”, never as “you rank #2 in ChatGPT”. Directional, reproducible, honest.

05Accuracy audit

Statements assistants make about you are compared with the facts you supply, so you can see where a model is confidently wrong.

Platforms covered

ChatGPT · live
Gemini · live
Perplexity · live
Claude · experimental
0

AI visibility

across repeated samples

Mentioned72%
Cited as a source38%
Positive sentiment61%

Illustrative sample — your numbers come from your own tracked queries.

For the metric definitions see AI visibility tracking, and for what to change once you have the data, read the nine-lever playbook.

See the method applied to your own brand.

Run a scan