Evidence first. No invented scores.

Every check runs five buyer-style prompts, such as “best tools for X” or “X vs Y”, on two live AI search engines: ChatGPT Search and Perplexity. Each prompt runs three independent times. That is 30 observations per check, collected through Bright Data against the live products rather than replaying model training data.

The report keeps the raw material visible: answer text, citation URLs, the model when reported, and the timestamp for each run. Mention and citation frequencies are counted from those observations.

AI answers vary between runs. A precise-looking number built from robots.txt, schema markup, or a simulated answer can hide the variance that matters. Repeating each prompt and attaching the underlying answers lets you inspect what the engines actually said.

A result is a measurement of this prompt set, engines, and time window. It is not a promise of future mentions, citations, or search position.

Each live check costs money to collect. The free tier gives makers three full checks per account without a credit card. The paid plan is for teams and makers who need up to 10 on-demand checks per calendar month; every check keeps the same 30-observation method.

Reports expose the raw answers, citations, model details, and timestamps, with report links available for seven days. See the pricing page for the exact offer and cancellation boundary.

A low mention rate is not a verdict. It is a prompt-level to-do list. Review the 20-point GEO checklist and the AI search optimization guide before deciding what to change, then run another on-demand check when you have a concrete hypothesis to test.