After this read you can decide whether your first job is to become known at all, or to correct facts the models already repeat.
We ask the assistants about your brand by name, with no website supplied, and compare what comes back to what is true. Sometimes nothing comes back at all. More often something does, and it is out of date: an old address, a parent company you left, a product line you stopped making. Those facts do not expire on their own. They get repeated.
Example result
Example. Real measurements, brand anonymised.
Brand asked by name on 12 August 2026, with no website supplied. Models: ChatGPT, Gemini and Claude. Unit: count of models repeating each fact, out of 3 models, over 3 runs each.
What this measures
INPUTS
Your brand name and, where you declare one, your display name. No website is supplied.
MODEL SET
ChatGPT, Gemini and Claude, each asked the same question in the same words.
REPEATED RUNS
3 runs per model, so a single lucky answer cannot set the reading.
OUTPUT
One verdict per model, plus every repeated fact with the count of models that returned it.
What you receive
KNOWN
Named and described correctly without prompting.
KNOWN, FACTS WRONG
Recognised, but the details returned do not match reality.
UNKNOWN
No reliable recall. The assistant cannot describe you unprompted.
Next action: correct the facts that recur across models first, then read the pages that carry them in Content.
Limitations
Assistants answer differently on identical prompts. We ask each model several times across runs and report what recurs, not a single answer.