Frequently asked question · Fix What AI Says
How will we know when each engine has dropped the old claim?
Because the prompts that surfaced it keep running every day, and each answer is kept in full. The accuracy log records the day the wording changed for each engine and each country, so you see that Perplexity dropped the old price in week two, Gemini in week four, and Copilot has not yet. Nobody promises a date, because the engines update on their own schedules and a memorised answer waits for a retrain. What you get instead is the record: what each engine said yesterday, what it says today, and the list of what shipped in between.
What the log holds for each claim
When a person on our team logs a wrong statement, the entry carries the prompt that surfaced it, the engine, the country, the exact wording, and the sources the answer cited. From that day on, the same prompt runs daily on that engine and in that country, and every answer is kept in full. The log marks the day the wording changes, so the moment the old price, the closed office, or the "legacy option" line leaves the answer is recorded as a date, not as somebody's recollection.
Because the record is per engine and per country, the same claim can be closed on one row and open on another. Perplexity in Brazil drops it in week two; Gemini in the United States drops it in week four; Copilot still repeats it. That is a normal shape for the log to take, not a sign that something failed.
Why the dates spread out
Each engine picks up a change on its own schedule. A page you fixed can be recrawled within weeks. A directory listing or a publication changes when its owner acts on the request, or never. A claim the model memorised in training has no page to recrawl and waits for the model to be retrained. We tell you which of those you are looking at for each open row, so an engine that has not moved is not a mystery.
Reading it beside what shipped
The monthly briefing puts the log next to the list of what went live in between: the rewrite, the redirect, the correction request, the new page. When a row closes, that list is what explains it. When a row stays open, the source that engine is still citing is what says what to do next. The record also carries the movers for the window, the prompts, sources, and competitors that gained or lost, so a fixed claim is read in the context of everything else that changed.
Answered on
Fix What AI Says
AI engines name your brand and get the facts wrong. We trace each claim to the page it came from, fix it, and track the day each engine drops it.
More questions about Fix What AI Says
See what AI is already saying about your brand.
A free audit shows you exactly where you stand across ChatGPT, Gemini, Perplexity, Copilot, Claude, Grok, and AI Overviews, and what it would take to close the gap.