When a potential customer asks ChatGPT, Claude, Perplexity, or Google's AI
answers for "a good [what you do] near me" or "a company like X", one of two
things happens: your business is in that answer, or it isn't. That channel is
quietly taking share from the old "search a list, scroll, click" path, and most
small business sites have never been checked against it.
You can check where you stand in about an hour, without hiring a
$2,000-a-month agency. Here is the method I use.
1. Run the same prompts a buyer would
Open a fresh chat in ChatGPT, Claude, Perplexity, and Google's AI mode. Ask each
one, in a buyer's words, not yours:
- Recommend a few [category] businesses in [city]. For each, list the name, what they're known for, and their website.
- I need [specific service]. Which local providers should I look at, and why?
- Compare [your business name] with two competitors for [use case].
Score each answer 0-3: 3 = you're named with an accurate description and a
correct link, 0 = you're absent or misdescribed. Do this for six prompts and
total it out of a fixed maximum so you can compare month to month. If you score
low, that's not a verdict, it's a baseline.
2. Check the things these systems actually read
AI answers are assembled from your site plus third-party sources. The gaps that
keep small businesses out of them are boring and fixable:
- Identity is inconsistent. Your business name, address, and phone differ between your site, Google Business Profile, and directories. Pick one exact form and make everything match.
-
No structured data. Add
LocalBusiness/Organizationschema with your name, location, services, and hours. It's a few lines in the page head. - Content isn't answer-shaped. These models favour pages that directly answer a question in the first paragraph. A page titled "Our Services" with a paragraph of adjectives gives them nothing to quote. "How much does X cost in [city]?" with a real answer does.
- No third-party corroboration. If only your own site says you exist, you're a weak citation. Reviews, a clean directory listing, a mention in a local round-up - each one raises confidence.
-
The page can't be crawled. Text baked into images, content that only loads
after a click, a
robots.txtthat blocks the AI crawlers. Check your robots.txt forGPTBot,PerplexityBot,Google-Extendedand decide deliberately whether to allow them. - Thin or missing basics. No clear list of services, no service-area page, no "about" with real specifics.
Walk your site section by section and mark each item Pass / Partial / Fail.
3. Turn the failures into a ranked list
Don't fix things in the order you found them. Score each failed item by impact
(how much it plausibly affects whether you're cited) times effort (how long the
fix takes). Do the high-impact, low-effort ones first: identity consistency and
schema usually top the list; a content rewrite is higher effort and comes later.
4. Re-test and keep the score
After each batch of fixes, run the same six prompts again and compare the total.
That loop - measure, change the top item, re-measure - is the whole method.
Whether an assistant names you depends on many factors outside any checklist, so
treat the score as a trend line, not a promise.
I put the full version of this into a small kit ($19): a 40-point AI-visibility
audit checklist, a prioritised fix worksheet with a default impact guide, and
the 6-prompt test with a scoring rubric and a place to track it over time. It's
a method and a set of templates - no ranking, citation, or lead-volume outcome
is promised. If it's useful:
AI Visibility Audit Kit.
There is also a small free tool that turns your business basics and a few
common questions into a labelled FAQ skeleton plus copy-paste FAQ schema, so
you have a first draft to fill in:
AI-Ready FAQ Starter.
It runs entirely in your browser and promises no ranking or traffic outcome.
Made by the Revenue Operator project. These materials were drafted with AI
assistance under human direction.
Top comments (0)