Baseline Measurement

A baseline measurement is the first reading of a metric, taken before anything is changed, so that every later reading has something to be compared against. In AI visibility it is the first run of the question panel: which assistants name the business, on which questions, in what position, before the site is touched.

What is frozen at the baseline

Four things: the questions, the assistants they are put to, the scoring, and the record. The questions are the ones a customer of that trade in that place would ask, and once written they do not change. The assistants are named. The scoring is fixed in advance, so a mention, a recommendation and a first-position mention each count the same way in month six as in month one. And every answer is kept verbatim, because a score that cannot be traced back to the sentence that produced it is a claim, not a measurement. Change any one of the four later and the comparison breaks. The methodology page sets out how this site fixes them.

Why one reading is a baseline and not a rate

Assistants do not give the same answer twice. A single run tells you where a business stood on one day; the second run tells you whether it moved. The nearest thing this site has to a baseline for small businesses as a group is a one-question probe of 9,240 of them, put to Claude with retrieval in September 2026: 14.9% were named. That is a Claude number, from one question per business on one day, and it describes the sample rather than any business in it. A baseline for one business needs the full panel, and the date is part of the reading. The probe is described in what 9,483 small business websites serve to an AI crawler.

Baseline and benchmark

A baseline is your own earlier reading. A benchmark is a reference from outside: a competitor’s figure, a sample average, a fixed test. The two are often used as if they were one word. A business can be above the benchmark and below its own baseline in the same month, and both statements are useful.

A common question

Why is one run of the question panel not enough to state a business's visibility?

Assistants do not give the same answer twice, so a single run records where a business stood on one day and no more than that. It becomes useful once a second run shows whether the position moved, which is why the questions, the assistants they go to, the scoring and the verbatim record are all fixed before the first reading is taken. Change any one of them later and the comparison breaks. The date belongs to the reading rather than sitting beside it as a footnote.

Sources

The probe figure is this site’s own, computed from its audit records; one assistant, one question per business, one day; sample drawn per metro and trade. Scoring: the methodology page.