Run and read a scan

A scan is a dated measurement: at this moment, across the engines you picked, here is what AI answers say in your space. Two screens follow one another — the one that launches, and the one that reads the result. The first is three decisions; the second takes a little method so you don’t over-read it.

Before you launch

• The brand is read-only, taken from your confirmed brief. “Edit” sends you to Réglages > Ma marque, and you come back here to launch.
• You tick the engines to measure. The list is the engines included in your plan, all selected by default: there is nothing to enable beforehand, and you cannot tick an engine your plan does not cover.
• You can also select up to 5 products. The scan then runs in product mode and the result is broken down product by product rather than for the brand alone.

Reading the result

• While the scan runs, a status badge and live refresh keep you posted: leave the page open, or come back later.
• “Overall mention rate”: the share of answers where your brand comes up, with a confidence interval and the gap against the previous comparable scan.
• One card per engine. When an engine could not be measured, the card reads “not measured”, never a 0 — a missing measurement is not a missing brand.
• “Signal reliability” and “Measurement coverage” tell you how much weight the headline figure can carry.
• Brand mode also shows “Mention rate by AI engine”, “Share of voice in AI answers”, “Brand sentiment”, and up to three verbatim excerpts under “What AI says about your brand”.
• Product mode shows a podium, “Visibility by product”, a share-of-voice donut and “Competitors AI recommends”.

“Export CSV” is a real export: semicolon-separated, UTF-8, and it deliberately includes the rows that failed. You get the full audit trail of what was measured and what was not, rather than a cleaned-up version. A finished scan cannot be re-run or deleted from this screen: to measure again, you launch a new one.