In this article
To check whether AI assistants recommend your business, ask ChatGPT, Perplexity and Google’s AI Overviews the questions your customers ask — recommendation, comparison and cost queries for your category and city — and log who gets named. Twenty minutes gives you a baseline share-of-voice picture that most of your competitors have never measured. Here is the exact protocol, and what to do with the result.
The 20-minute protocol#
- Write five buyer prompts: "best [your service] in [your market]", "[service] cost", "[you] vs [competitor]", "who should I hire for [problem]", "is [your brand] good".
- Run each in ChatGPT (with browsing), Perplexity and a Google search that triggers AI Overviews.
- Log three things per answer: brands named, sources cited, and what is claimed about you — correct or not.
- Repeat monthly, same prompts. The trend is the metric.
Reading the results#
Absent everywhere? The engines cannot verify you exist — usually thin content, no structured data, or a site AI crawlers cannot read. Named but with wrong facts? Your entity data is inconsistent across the web: directories, LinkedIn and your own pages disagree about services, locations or pricing. Competitors cited via a comparison article? That article is now the market’s sales rep — and the strongest argument for publishing your own citable comparison content.
The fixes, in order of leverage#
- Verify AI crawlers are not blocked in robots.txt and content renders without JavaScript.
- Publish answer-shaped pages: your prices, comparisons, FAQs — the exact material engines lift into answers.
- Fix entity consistency: identical business facts on your site, directories and social profiles.
- Add original numbers — case results, benchmarks, real prices. Engines prefer citing the primary source.
- Get into Bing Webmaster Tools; ChatGPT’s browsing runs on Bing’s index.
You cannot manage a channel you have never looked at. Twenty minutes of prompting tells you more than a year of assuming.
Frequently asked questions#
AI answers vary between runs — is measuring even meaningful?
Yes, in aggregate. Individual answers fluctuate; consistent presence across prompts, engines and months does not. Treat it like rankings: trend and share of voice, not a single screenshot.
How fast can absence be fixed?
Faster than classic SEO. Engines favor fresh, structured, fact-dense content — sites that publish citable pages and fix crawlability have appeared in answers within weeks in our projects.
Is there tooling, or is it all manual?
Dedicated AI-visibility trackers exist and mature monthly, but the manual protocol is the right start: it costs nothing and forces you to read what engines actually say about you. Automate once the baseline exists.