4%
of local queries get the same answer from all four major AI engines. One engine's result says little about the others.
Yext, 491,972 local queries, Feb to Aug 2026
18%
of consumers would contact the assistant's pick without checking further. 43% then checked Google.
BrightLocal, July 2026, n=1,227
6 of 6
distinct businesses named across two metros in most of the verticals we probed. No one owns the pick yet.
Picked by Agents probe, 32 runs, Sept 2026
Two ways to measure AI visibility
Dashboards such as Profound, Scrunch and Athena run large sets of prompts across AI engines and count how often a brand appears, where it sits in the answer, and how the answer describes it. For a national brand with thousands of product questions, that breadth is useful.
A local service business has a handful of questions that pay the bills, asked from specific ZIP codes. For those, a mention rate is the wrong unit. You want to know, question by question, whether the assistant told the customer to call you.
What a pick report shows
- 01
The question, word for word
Written the way your customers ask it, with a place and a need, and agreed with you before we start.
- 02
Who the assistant named
The businesses it recommended, in order, in each assistant we ran. We describe the competitors to you and never publish their names on your report.
- 03
Why it passed on you
The assistant's own reasoning, saved from the run. No price found, no sign you cover that ZIP, a phone number that did not match the profile.
- 04
What we fixed and what moved
Each fix tied to the reason it answers, and the same question re-run the next month.
Why reasons matter more than scores
A score tells you that you are losing. A reason tells you what to change. When an assistant writes that it chose a competitor because they list a starting price and offer same day visits, the fix is a sentence on your service page, not a quarter of content.
“A visibility number that goes up while the assistant still names someone else is not progress.”
Three numbers, never blended
Picks are the leading measure. Behind them we report AI referrals we can observe in your analytics, the customers who told you an assistant sent them, and our own estimate of new business. Each stands alone, so you can see which is solid and which is an estimate.
Many AI referrals do not show up as referrals at all. A customer reads the answer, then searches your name or calls from the profile. That is why we ask your team to log how customers heard about you, and why we never present an estimate as a count.
Typical AI visibility dashboard vs Picked by Agents
| Typical AI visibility dashboard | Picked by Agents | |
|---|---|---|
| Unit measured | Mentions across a large prompt set | The pick on each real buying question |
| Prompts | Generic category prompts, often national | Your customers' questions from the ZIP codes you serve |
| Output | Share of voice, sentiment, position charts | Who was named, the assistant's reasons, the fix list |
| Who does the work | You or your agency, from the dashboard | We fix what you own and re-run the question |
| Business results | Estimated traffic | Observed referrals, customer reports and our estimate, kept apart |
Questions we get asked
- What is AI visibility?
- AI visibility is how often and how well a business appears in answers from AI assistants such as ChatGPT, Gemini and Google AI Mode. Most tools measure it as a mention rate across many prompts. We measure the part that pays, whether the assistant names you when a customer asks who to hire.
- Should I use Profound, Scrunch or Athena instead?
- If you are a large brand that needs to watch thousands of prompts across products and markets, a dashboard is the right tool. If you are a local business with a few questions that decide your month, you need those questions run, the reasons read, and the fixes made. That is what we do.
- How do you avoid lucky results?
- Assistants vary run to run. We run each question repeatedly, across the assistants that matter, from the places you serve, against a baseline agreed in writing. One good answer never counts as a pick.
- Can you track AI traffic in my analytics?
- We tag the referrers that assistants pass along and set up call tracking tags for AI sourced inquiries on our Picked program. Much AI influence arrives as a name search or a call from your profile, so we pair it with what customers tell your staff.
- Is the first report really free?
- Yes. It runs your customers' questions, shows who got named and why, and lists what stops the pick. You see where you stand before you pay anything.
Measure the pick, not the mention.
Get a free pick report on your customers' real questions, with the assistant's reasons saved for every answer.
Or write to Greg at greg@gregmiller.ai.