Most businesses have no idea how AI assistants describe them. They've never asked. Meanwhile a meaningful share of their buyers are forming first impressions inside exactly those tools.
This is measurable. It takes about an hour a month and no special software.
Build a fixed prompt set
The discipline is that the prompts don't change. Same questions every month, so movement means something.
Fifteen to twenty-five prompts across four types:
- Category discovery — “best Shopify development agency in Florida”.
- Problem-led — “my Shopify store is slow, who can fix it”.
- Comparison — “how do I choose between agencies for AI consulting”.
- Branded — “what is [your company]”. This one catches misinformation.
Run them consistently
Cover ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews. Use a fresh session or temporary chat each time — personalization from your own history will flatter you and corrupt the measurement.
Record five columns
- Mentioned? Yes or no. This is your headline metric.
- Position. First name, or sixth in a list?
- Description accuracy. Correct services, correct location, current?
- Sources cited. The most actionable column by far.
- Competitors named. Who's winning and how are they described?
Reading the results
Never mentioned anywhere — an entity problem. These systems can't confidently resolve who you are. Fix schema, the About page, and consistency across the web first.
Mentioned but described wrongly — stale or contradictory information somewhere. Find the source and correct it.
Mentioned on some platforms only — they weight sources differently. Look at what the platforms citing you have in common.
Competitors cited from sources you're absent from — the clearest to-do list you'll get. Those directories, roundups, and review platforms are where the answers come from.
A realistic benchmark
For a small regional business, appearing in a third of category prompts is solid. Branded prompts should be near total — if an assistant can't describe you accurately when asked by name, that's the first thing to fix.
Why bother now
Because it's early. The sources these systems lean on are still winnable, and most of your competitors haven't looked once.