Evidence before interpretation.

This is an observation recorder.

The tool analyses answer text and source URLs supplied by the user. It does not query, scrape, impersonate or independently reproduce ChatGPT, Google AI Mode, Gemini, Perplexity, Copilot or another answer engine.

Every report therefore says observer-supplied evidence. A saved result proves what was entered at that time—not that every user would have received the same answer.

What is frozen.

Engine

The answer surface selected by the observer.

Exact prompt

Stored up to 500 characters so prompt variants stay distinguishable.

Target

A normalised domain, with www. removed.

Answer

Observer-supplied text, stored up to 20,000 characters.

Sources

Up to 40 valid HTTP(S) URLs, deduplicated in supplied order.

Time

Observer-supplied time when valid, otherwise save time.

Simple checks, fully disclosed.

A target citation is detected when a supplied source hostname equals the target domain or is its subdomain. Citation position is its first position in the supplied list.

The brand mention detector derives a literal phrase from the first target-domain label: hyphens and underscores become spaces, matching is case-insensitive, and the phrase must be at least three characters. This intentionally avoids undocumented entity guessing.

A compact fingerprint—not a visibility index.

If the target is cited below first position, the position component falls by two points per place until it reaches zero. Scores are capped at 100. The score only summarises this one supplied observation.

What the report cannot establish.

• It cannot verify that pasted wording or URLs are complete or authentic.
• It cannot account for personalisation, location, account history or live model changes.
• It cannot detect brand aliases that differ from the target domain stem.
• It cannot infer traffic, citation share, market share, sentiment, causation or future visibility.
• It cannot tell whether a citation supports the answer accurately.

Use repeated, consistently collected observations to describe a trend. Preserve screenshots or exports when a result needs independent evidentiary support.