Short answer: the report needs the method before the numbers, two rates rather than one score, the evidence behind any figure, and an explicit record of what was not measured. Drop the fourth and the first three stop being trustworthy.
Section 1 — Method, first, not in an appendix
The question set and its version. The engines and the markets. The collection window. How competitors were detected. Whether brand-named questions were excluded. This is four sentences and it is what separates a report from a screenshot — because every question the reader has later is answered here.
Section 2 — Presence, split into two rates
Mention rate and recommendation rate as separate columns, never combined. A brand mentioned in 40% of answers and recommended in 4% has a corroboration problem. The reverse pattern is an attribution problem. One number cannot distinguish them, and those two situations need completely different work.
Section 3 — Citations, counted honestly
Only sources marked in the answer body. The retrieval candidate pool an engine assembles is much larger than what it cites — counting the pool inflated our own figures by two to four times before we caught it. Report rank among all cited sources rather than a share percentage: "fifth of 109 sources" is interpretable and stable; "5% citation share" moves for reasons unrelated to anything you did.
Section 4 — What was not measured
| Situation | Correct rendering | Common mistake |
|---|---|---|
| Engine timed out | Blank, with a note | 0%, which turns a flat trend into a decline |
| No run scheduled that week | Gap in the series | Interpolation, which invents data |
| Question failed to parse | Excluded, and said so | Dropped silently, which raises the average |
The property that makes it useful
Every number should be openable. Click a rate, see the questions behind it; click a question, see the answer text, the engine, the date and the cited URLs.
Without that, a client asking "why did this drop?" gets a guess, and you cannot tell a detection bug from a real change. In this category you will have both, regularly.