Two numbers appear on this site and they count different things. Both are below, with the counting basis behind each one.
Distinct hostnames with a stored visibility score, including partial or otherwise qualified results; excludes dead, error, failed, eval and test sources; counted 23 September 2026.
The exact figure on that date was 4,429 hostnames across 6,257 scored scan rows. Of those, 920 hostnames have at least one result we flagged as degraded, for example because the market or locale read was uncertain, or an engine did not answer, and 45 have at least one flagged as suspect. Both are kept in this number, because the scan happened and produced a reading.
This is a dated snapshot. It does not move on its own. When we re-count it, the date changes with it. We round the published figure down to the nearest hundred.
The narrower measure, and the live one. A website counts as measured when a visibility score was recorded against a real buyer question for its category. This count also excludes dead, error, failed, eval and test sources, and records without a source.
Two kinds of reading that the scanned number keeps are left out here. The first is a self-match, where the category we read was the business's own brand, so the question answered itself. The second is a reading where our namesake guard tripped, meaning the business shares a name with something larger and we cannot be sure the web footprint we measured is theirs. That second kind still counts in the business's own trend, where it is only ever compared against itself. It stays out of a platform total, where it would credit us with a measurement of somebody else.
Readings we flagged for a milder reason, an uncertain locale or an engine that did not answer, still count as measured. A hostname counts if it has at least one eligible row.
At the moment each number was counted, every measured website had also been scanned, so measured is the narrower of the two. The scanned figure is frozen at its count date and this one is live, so as we keep measuring, the live measured total can eventually pass an older scan snapshot. When that happens it is the two dates moving apart, not the definitions disagreeing.
We count hostnames, not businesses. There is no owner or company field in this data, so one business with eight websites counts eight times. Anyone quoting these numbers as a count of companies is reading them wrong, and that includes us.
We also do not claim these websites are a sample of anything. They are the sites people asked us to check, plus the sites we added to build category coverage. They are not drawn at random and they are not representative of any market.
The public AI Visibility Index and its percentages use their own stated cohort. The websites-scanned snapshot is not the size of that Index, a count of free-check customers or the number of websites scanned each night.