Vouch

Customers ask AI what’s good nearby. We grade the answers.

00:00 idle
A block of storefronts drawn in ink on paper; three shop windows lit

The streetthe shops an engine names when asked what is good nearby

The street appears when the first probe returns.

Lit = an AI engine cited this shop when asked. Plate = Vouch monitoring is active. Neither changes the other.

The readingscore computed in code from ground-truth diffs, never graded by a model

A surveyor's desk seen from above: a street map, a benchmark plate, a dial gauge, calipers, a pencil and a blank notebook
No audit yet. Run one to see what AI says about a business.

The non-determinism probeone engine, one prompt, asked twice

Not yet run.

First ask
Second ask

Source overlap is measured once the same prompt has been asked twice.

Agentsa human approves before anything leaves the building

Cost of this run

Costs are reported when the run completes.

Engines probed: Claude, with web search Ground truth: Google Places (New) We measure answers; we do not claim to change them. Instrument idle · no reading yet run id not assigned