Define the population the report describes
Record the approved question set, period, market, provider and observed surface. A report about a fixed ChatGPT question set should not claim to represent all AI search or all users.
Measure AI visibility before filling the template. Report answer-level counts consistently: one answer with three brand mentions still contributes one to the answer-level mention count.
Calculate coverage before interpreting a rate
Coverage is valid answers divided by planned answer runs. Mention rate is valid answers naming the brand divided by valid answers. Owned citation rate uses answers containing an owned source link as its numerator and answers with complete citation evidence as its denominator. Report missing citation evidence separately.
Illustrative example: 80 planned runs, 74 valid answers with complete citation evidence, 28 answers mentioning the brand and 12 citing an owned page produce 92.5% coverage, 37.8% mention rate and 16.2% owned citation rate. These fictional counts demonstrate the arithmetic; they are not customer results. A zero denominator is reported as unavailable, not zero.
Attach the original evidence
Keep prompt IDs or exact question text, original answers, source URLs, collection dates and the meaning of each finding. Distinguish a citation to the client’s site from an independent source. Understand citations versus mentions.
The template calculates a draft from your inputs. It does not collect answers or authenticate the evidence you paste.
Describe changes and unresolved work
Add the action owner, acceptance criteria, status and recheck date. Explain any change in questions, market, aliases or collection method before comparing periods. An observed movement after an edit does not prove causation.
Use client reporting to see how evidence and assigned work stay connected in a project. Keep technical completion, observed outcomes and future hypotheses in separate paragraphs.