Quarterly AI visibility audit: the 7 questions that catch blind spots

Run the same prompt set every quarter, or your AI mentions become impossible to compare. Keep it simple: brand name, product names, category queries, problem-based queries, and “best X for Y” searches. Test both short and long phrasing, plus one query with a competitor nearby. If the answer changes, log the exact wording, the model, and whether your site was cited, paraphrased, or ignored.

Use a sheet with these columns: date, prompt, model, country or locale, result type, cited URL, mention sentiment, and notes. Add a separate tag for “new,” “lost,” or “unchanged.” That lets you spot patterns fast: pages that earn citations, pages that get summarized without attribution, and pages that never enter the answer set. The goal is not volume; it is repeatable visibility.

Questions to include:
— Does the model name the brand without help?
— Does it cite the homepage or a deeper page?
— Does it prefer a list, glossary, comparison, or how-to page?
— Does it mention competitors first?
— Does it confuse the brand with another entity?
— Does it give a direct answer or only a generic one?
— Does it use your wording, facts, or structure?

After each audit, look for one action only: improve one page, add one comparison block, or tighten one entity signal. If you track the same prompts every quarter, you stop guessing and start seeing which pages AI systems actually trust.