We show our working

Answer engines moved from novelty to default in under two years. Buyers now ask an assistant which provider to choose, and get one answer with a few names in it. For the people accountable for a brand, that is an enormous shift with almost no instrumentation behind it.

The first wave of tools filled the gap with a single visibility score. The problem is that the underlying thing is not stable enough to justify a decimal place, ask the same question twice and you can get two different sets of brands. A precise-looking number built on that is not measurement, it is decoration.

Report the range. Where the honest answer is a band, we publish a band, along with a confidence trend that tells you how settled it is.

Show the evidence. Every figure expands into the prompt we asked, the engine's own words, the timestamp and the engine used.

Name the limits. Sampling is a clean session, not a personalised one. Location and device are settings we choose, not facts we observe. Two engines are captured from the live surface rather than an API. All of that is written down in public.

No gamification. No leaderboards, no streaks, no badges. The people using this are making budget and messaging decisions.

We sample six engines on a rolling schedule with fixed market and device settings, record every response in full, and version the methodology so a change in how we measure is never mistaken for a change in how the brand is described. When a model version is pinned, it is recorded with the response; when it changes, we treat that as a methodology event and say so.

Where an engine simply did not serve an answer, we record "not served" rather than inferring a zero.

The methodology explains what counts as a mention, how ranges and confidence are built, and lists every caveat we know of, including the ones that are inconvenient for us.

Seenford is a trading name of FranklyCo Ltd, registered in England and Wales, company no. 17392627. Contact: hello@seenford.com.