SEO & AI visibility
How to measure AI visibility reproducibly
Keep prompts and evidence consistent across engines and time. Record the provider, model, surface, country and date with every run, repeat runs before drawing conclusions, and treat API answers and consumer answers as different surfaces. One prompt is a sample; a prompt group tracked over weeks is a trend.
Choose customer questions
Use questions a buyer could ask without knowing your brand. Keep branded questions in a separate part of the prompt set. Category, country, buyer type and budget all affect results. Write down the exact question rather than paraphrasing it when you run another engine.
Capture the context
Use your own ChatGPT, Claude or Gemini account and record the actual answer. Save displayed model information when available, sources, the run number and the date. A fresh conversation reduces conversation-history effects but does not guarantee an impersonal or region-neutral result. Do not silently replace a failed engine with another model.
Compare like with like
Keep consumer answers separate from API measurements. Repeat the same prompt set after a change and inspect the underlying excerpts. A higher mention rate in a small sample is a reason to investigate, not proof of permanent visibility growth. Unknown recommendations and missing responses should remain visible in the report.
See the measurement methodology for product-specific coverage and limitations.