The 8 checks
- Category question. Ask the question a buyer asks before they know the vendor list. Record whether the brand appears at all.
- Comparison question. Ask “X vs Y” with a named competitor. Record who is cited as evidence.
- Money question. Ask about pricing or “is X worth it”. Pricing pages are frequently cited — or not.
- Citation owner. For each answer, list the domains cited. A third-party comparison page winning the citation is the finding.
- Position. If the answer lists vendors, record where the brand sits. Position moves faster than mention rate.
- Run variance. Repeat each question on different days. Brand sets change between runs; one run is a sample.
- Competitor set. Record who appears next to you. The engine is telling you which set you belong to.
- Day-60 re-run. Re-check after content changes. Without a dated re-run, improvement is an assumption.
What makes an audit credible
Dates on every answer, the exact question text, the engine and the mode, and the citation list. If any of those is missing, the number cannot be checked. The same rule applies to tools: a tracker that cannot show its raw answers is a dashboard, not a measurement.
Our own method follows this: three runs per question across three days in the paid diagnostic, and a published free one-run sample. The variance study shows why: the same 15 questions, re-asked three times, kept an identical brand set in only 6 of 15.
Free checklist (start here)
- Write down 10 buyer questions: 4 category, 3 comparison, 3 money.
- For each, run the free check and copy the answer, the mentioned brands and the cited domains into a row with the date.
- Mark rows where the brand is mentioned but the citation goes elsewhere — that is the shortlist to fix first.
- Repeat the same questions in two days. Compare brand sets. Pick the pages that won the citation consistently.
Related
ChatGPT brand monitoring: the 5 numbers · ChatGPT brand tracker: what to record · Rank tracking vs AI answer tracking · AI visibility index (35 brands)