Corrections
Report an error in Parse
Found a mistake in Parse? Send us the page and tell us what looks wrong.
Tell us what happened
Use Report an issue on the page, or write to [email protected].
Send the page link, what looks wrong and anything that helps us check it. You do not need to diagnose the problem first.
A person reads every report. If we confirm an error, we fix its cause. The correction appears when we next publish the affected data.
What we correct
A brand in the wrong market. Two listings for the same company. A description or website that belongs to someone else. A number that does not match the answers behind it. A mistake in our research or writing.
What the record preserves
An AI answer can be wrong even when Parse records it correctly. We preserve what the AI actually said.
We correct errors in how Parse collects, identifies or counts those answers.
Research corrections
October 6, 2026 · How many prompts do you need to measure AI visibility?
The previous version overstated how certain its estimates were and suggested a question count that could apply to any brand. The analysis did not establish either claim.
We removed those conclusions and rebuilt the study around a published dataset covering four markets. The revised report measures sampling error within that dataset; it does not claim to establish a universal question count.
Our validation studies explain what we tested, which errors we found and what remains uncertain. Each report’s updated date and supporting downloads identify the version you are reading.