Do AI assistants recommend your product? How to measure AI visibility
You find out by asking them what your buyers would ask, many times over, and counting. Run a fixed set of prompts on several assistants, repeat each one, and track the share of answers that name you and the pages they cite. Start with questions about your own product, then how-to questions your docs should answer, then your category. Add the free reports the search engines now publish. This is how we’re starting to measure it ourselves, and we don’t have results yet.
What is AI visibility?
It’s how often AI assistants name you, and cite your pages, when people ask questions you should come up for. You’ll also see the work called GEO, for generative engine optimization, or AEO, for answer engine optimization. The tools that track it report the same few numbers: visibility, the share of answers that name you; share of voice against others in your category; citation share, which of your pages get cited; and sentiment.
Why isn’t one check enough?
Because the answers change nearly every time. In a SparkToro study, 12 prompts asking for brand recommendations were run 60 to 100 times each on three AI tools, 2,961 runs in all. Two of the three returned the same list of brands less than once in a hundred runs. How often a brand appeared was steadier. So a position in an AI answer means little. What to track is the share of runs that name you, and that takes many runs per prompt.
Which questions should you track?
Three groups, in this order:
- Questions about you. “What is your product?”, in your product’s name. Does each assistant know you, and is what it says right? A wrong answer here is the first thing to fix, usually by making the right answer easy to find on your own site.
- How-to questions your docs should answer. For us, questions like “how do I make an AI agent ask a person before a tool runs”. An answer that cites your docs is the clearest sign your pages are being read.
- Category questions. “What’s the best tool for…” A new name rarely shows up here at first, because the answers tend to favor brands with years of mentions. Track it for the trend, not for this month’s score.
About 30 prompts across the three groups is enough to start.
How do you run the prompts?
Through the assistants’ APIs, with web search turned on, on a schedule. Run each prompt 5 to 10 times on each of 4 or 5 assistants, once a month, and keep every answer in a spreadsheet: the prompt, the assistant, the date, whether you were named and which pages were cited.
For example: say Acme, a made-up online store, tracks “what’s a good waterproof jacket for day hikes”. It runs that prompt 10 times on each of 4 assistants, 40 answers in all. If 6 answers name Acme and 2 cite its pages, Acme’s visibility on that prompt is 15%, and next month’s 40 answers show whether it’s moving.
Know the limits. An API can answer differently from the app people use, and neither sees what a signed-in person with history and memory sees. The study’s authors left open whether API calls vary the same way. Nobody outside the assistants knows how often each question is asked. Treat the numbers as a trend, not a ranking.
What do the search engines tell you for free?
More this year than before, and still not everything:
- Bing Webmaster Tools added an AI Performance report in February 2026: how often your pages are cited in Copilot and in Bing’s AI summaries, which pages, and the phrases the AI used to find them.
- Google Search Console added generative AI performance reports in June 2026, open to every site since August: impressions in AI Overviews and AI Mode, by page, country, device and date.
- Your own logs show which AI crawlers fetch which pages, and your analytics show visitors arriving from the assistants’ sites.
The chat assistants themselves publish nothing per site.
Do coding agents count?
Yes, if developers are your buyers. A developer’s coding agent reads documentation to build with a tool, and it may be the one that suggests the tool. It reads pages as plain text, so give every docs page a Markdown version and publish an llms.txt that lists them. Fetches of those files in your logs show coding agents reading you. More in When a coding agent builds your AI agent.
Where we are
Kindgi is an agent framework for production: your coding agent writes the tools, agents and flows in your own code, and Kindgi runs them and keeps every step on record. We’re starting to measure its AI visibility now, with the free reports first. We have no results to share yet, and a new name rarely shows up in category answers early on. This article is part of the work: we answer the questions people ask, and say why we write them. For coding agents, every page of our docs has a Markdown version, and llms.txt lists them.
How this was written: a coding agent drafted it from public research on AI visibility, including the sources linked above. We checked every claim, edited it and said yes. We publish articles like this partly because we want you to find us, including through AI assistants. More in How we write our posts.