Good point about one check being one sample. If it helps anyone turn this into a routine, here's a lightweight way to run it by hand each week:
Write 8-10 buyer questions in a sheet, phrased the way a customer would ask them, not how your marketing page words them.
Run each one in ChatGPT and Gemini with search on, in a fresh chat each time. Log three columns per engine: the brands named in order, the cited domains, and the date.
Run the whole list once a week for four weeks. Only treat a gap as real if it shows up in at least three of the four runs.
Pull the cited domains into a pivot and count them. The review sites, comparison pages and forums that show up repeatedly are your outreach list, since they feed the answer more than your own site does.
For each gap, ask the model directly: "Here is the answer you gave and the sources you cited. What does the cited page say about us that's missing or wrong?" Then fix the specific facts on your page.
Keeping the exact prompt wording identical between runs matters more than the number of questions.
Charlie from xSquads
Good point about one check being one sample. If it helps anyone turn this into a routine, here's a lightweight way to run it by hand each week:
Keeping the exact prompt wording identical between runs matters more than the number of questions.