How would you test whether AI assistants actually recommend a local business?
retained text
Everyone has a theory about how ChatGPT or Perplexity picks local businesses, but I haven't seen many repeatable tests. So here's how I'd set one up, happy to be told it's wrong. Write 10 or 15 questions the way a customer would type them, plus a reworded version of each. Keep the location the same every time. Ask each one three times, on different days, in ChatGPT, Claude, Perplexity and Gemini, and write down which businesses get named and which pages are cited. Then count how often a business shows up (say 4 runs out of 12) instead of judging one answer, because one answer is mostly noise. If you change something on the business side, like reviews, a listing or a page, change one thing and rerun the same questions a few weeks later. And open the cited pages to see what they have that yours doesn't. I don't have results to share, and as far as I know nobody publishes how these tools weight their sources, so I'd treat any single answer with suspicion. Has anyone here actually run something like this? What did you track?
document fetched 2026-10-29 00:45 · canonical https://reddit.com/r/digital_marketing/comments/1xhuzbn/how_would_you_test_whether_ai_assistants_actually