What is an AI Visibility baseline?
An AI Visibility baseline is the frozen record of how AI models answer about a brand before anything is published. It is measured across a fixed, versioned set of target prompts, in an anonymous session, with the date and engine on record. Without a baseline, no later result is demonstrable, all that remains is the impression that things improved.
Why the baseline has to come first
Because it cannot be reconstructed afterwards. AI answers are not publicly archived and they change with the index, the model and the date. If the first measurement happens after the first publications, the starting point is lost for good, and any number presented later becomes a claim with no counter-evidence.
It is also what separates measurement from sales narrative. An agency that never froze a baseline can always say results improved; one that did has to show in which questions they improved, in which they did not, and who showed up in its place in the ones that stayed closed.
What goes into a proper baseline
The target-question set, with a declared mix of category questions and branded questions, a set made only of the firm’s own name inflates the result and measures nothing. The engines tested, the date of each run, the number of repetitions per question, and a log of every brand cited, not only the client’s.
We also keep the text of the answers and the list of cited sources. That is what makes it possible, months later, to say not only that citations rose but what changed in the framing: whether the firm went from a passing mention to a linked source, for example.
Related terms
The baseline is the reference for citation rate, AI share of voice and the citation gap. It is the first deliverable of workstream 01 of the ReBo method, detailed in how AI Visibility is measured.
Frequently asked questions
Why does the baseline have to come before publishing?
Because AI answers are not archived and shift over time: once you publish, the earlier state can no longer be reconstructed. Without that record there is no way to tell a result from a coincidence.
How often is the baseline redone?
The baseline is not redone: it is the zero mark. What repeats is the measurement, on the same ruler, every month. When the question set needs to change, we create a new version of the set and declare the change in the report, preserving the old series.
How many questions go into a baseline?
Enough to cover the audience’s decisions and not so many that the measurement becomes impossible to verify by hand. What matters more than the count is the mix of category and branded questions, and the fact that the set stays frozen.