There is nobody to email
When an assistant states something false about your company, there is no listing to edit and no form to submit. The output was assembled from sources, so the only durable correction is changing what those sources say and waiting for the model to see it.
Find out what it read
Ask the same question several ways and look at the citations. Wrong claims usually trace to a specific origin: an outdated directory entry, an old press mention, a competitor’s comparison page, or your own site contradicting itself.
Your own site is the first suspect
More often than teams expect, the model is faithfully repeating something you still publish — an old service you dropped, a stale claim, a page describing a positioning you abandoned. That is the cheapest fix and the one a stale-fact audit is designed to catch.
Before blaming the model, check whether you are still publishing the thing it repeated.
Then make the correct version easier to find
Publish an unambiguous, specific statement of the correct fact on a page that owns the topic, and make sure your structured data agrees with your copy. Contradiction between the two is itself a cause of confident errors.
Get it corroborated
One page asserting a fact competes with several sources asserting the old one. Updating your listings, profiles, and anywhere you are described elsewhere is what actually shifts the balance — the slow, durable authority work that nothing else substitutes for.
Then re-check on a schedule
Corrections propagate unevenly and at different speeds across engines. Add the specific wrong claim to your fixed prompt set and watch it monthly, which is exactly what tracking how you are described is for.