Problem-first
The customer describes a symptom, not a service. The hardest family to win and the most valuable.
"my roof is leaking after the storm, who should I call in Broward?"
Methodology
How we measure whether AI recommends you
There is no API that reports your AI visibility. The only honest way to find out is to ask the engines the questions your customers ask, read what comes back, and count. That is all the scanner does — carefully, and the same way every time.
20buyer-intent prompts written for your category and market
5engines queried live, each with web access enabled
100real searches per scan, no sampling or extrapolation
~6minutes from pressing run to a finished report
Step one
The twenty questions
Six intent families, weighted toward the way people actually describe a need. Nothing generic, nothing keyword-stuffed — these are sentences, because that is what gets typed into an assistant.
"Best X in Y"
The classic local query. Answered almost entirely from third-party roundups and directories.
"best med spa in Fort Lauderdale"
Ranked comparisons
Head-to-head shortlists. Appearing here means the engine considers you a peer of the leaders.
"compare the top three personal injury firms near Plantation"
Price & rating filters
Qualified buyers with a constraint. Requires structured facts about you to exist somewhere public.
"affordable HVAC repair in Coral Springs with 4.5 stars or better"
Near-me
Proximity intent. Leans on map and business-profile data, so local signal problems surface here first.
"med spa near Las Olas open on Saturday"
Asked about you by name
The control group. Losing these is an emergency; winning only these means nobody discovers you.
"is Coral Ridge Med Spa any good?"
Step two
How each engine is asked
All five get the same twenty prompts. What differs is the door we knock on — and in every case, live web access is switched on so the answer reflects today.
Perplexity
Search-native API
Highest citation rate for local businesses, and the clearest read on whether third-party listicles carry your name.
Google AI Overviews
Scraped via SerpApi
Captured exactly as a searcher is served it. The highest real-world traffic impact of the five.
ChatGPT
Responses API · web_search
The default assistant for most consumers. Blends live results with a strong prior from training data.
Claude
Messages API · web_search
Conservative about naming businesses it cannot corroborate, which makes an appearance a strong quality signal.
Gemini
generateContent · grounding
Closest to the Maps and Business Profile ecosystem. Weak local data shows up here before anywhere else.
Step three
What gets counted
Three numbers, each defined narrowly enough that you can argue with it. That is the point — a metric you cannot audit is a metric you cannot act on.
Visibility
Were you named at all?
The count of answers that mention your business by name, divided by the answers we successfully received. Aliases you supply count as you.
Share of voice
How big a slice?
Your mentions as a fraction of every business mention across the whole scan. This is the number that says how crowded your category really is.
Cited as a source
Did they link you?
Answers that cite your own domain, not just your name. Being mentioned is good; being the source is what compounds.
Denominator integrity
If a request fails — a timeout, a rate limit, an engine having a bad afternoon — it is removed from the denominator instead of being scored as a miss. A scan that attempted 100 searches and completed 96 reports out of 96, and says so on the report. Inflating your score with someone else's outage would make the number useless.
Then we read it with you
A number on its own starts an argument. The report ends it, because every recommendation is tied to a measurement you can see on the page above it — which engine, which question family, which competitor, how many times.
Most of what follows is unglamorous: getting listed in the roundups the engines already trust, fixing structured data, earning two or three credible citations, and rescanning on a fixed cadence so you can tell a real change from a fluctuation.
How long until the number moves?
Directory and listing work can show up within a few weeks. Source-quality work — real editorial citations — takes a quarter or more.
Is this different from SEO?
It overlaps but it is not the same. Classic SEO optimizes a page to rank. Generative engine optimization makes your business the answer, which depends far more on what other sites say about you than on your own page.
Do the results change between scans?
Slightly, yes — these are generative systems. A one- or two-point wobble is noise. A trend across several weekly scans is signal, which is why the report charts it.
Do I have to hire you afterwards?
No. The first scan and the walkthrough are free, and the report is yours either way. Plenty of the fixes are things an in-house marketer can run without us.
Run it on your business.
Twenty minutes on a call, one free scan, and a number you can act on.