Signal · AI visibility
Top of one AI assistant? The other three may name someone else.
Two provider studies found that AI assistants often name different local businesses for the same question, and that asking again often returns a different list. A screenshot from one assistant on one day tells you about that assistant on that day. Our read: to know where you stand, ask the same questions on a schedule and write the answers down.
Published
Four assistants, one question
They rarely put the same business first
Yext compared four AI engines answering the same local questions. Each question all four answered counts as one local case, and in those cases they rarely agreed on who came first.
Of 491,972 local cases where all four engines answered the same question, how often they put the same business first
All four named the same business first4%
At least one put a different business first96%
Being first in one assistant says little about the other three.
One assistant, asked twice
Ask again, and the list changes
BrightLocal asked the same local questions more than once. The answers moved inside a single assistant, not only between assistants. Overlap compares the two lists as a whole; persistence asks whether a business named the first time was named again.
Reported by BrightLocal
20–33%
average overlap between one run and a repeat run
Overlap is the share of businesses named on both lists when the same question is asked twice. Across 200,085 localized searches.
Reported by BrightLocal
~50%
of businesses named once were named again later
About half of the businesses named in one run were still named in a later run. Across 200,085 localized searches.
What these studies don’t tell you
Both are provider studies with their own questions, engines, places and dates. They show the disagreement is real and large, not a forecast for your market.
The 4% is how often four engines agreed in that sample, not a probability for any one business. The overlap and persistence figures are averages; your own may sit above or below them.
Nothing here shows one assistant is more accurate or more worth chasing than another, or that any action wins a rank.
Read it carefully
Being named is a start, not a sale
There’s no single leaderboard across assistants. Even a good result answers a smaller question than it seems to.
Being named tells you
- This assistant mentioned you for this question, from this place, at this moment
- You made the short list it was willing to show someone
It doesn’t tell you
- That you were the one chosen from that list
- That other assistants named you, or will next week
- That anyone booked, walked in, paid or was served
- That your name, address, hours and category were read correctly
Nothing here measures customers. Keep AI mentions, enquiries, bookings and payments in separate columns, so a better mention is never reported as revenue.
What to do
Check it properly in four weeks
No new system needed: the same questions, on a schedule, written down. Four weeks is our suggested starting point, not a threshold either study set.
Write down five real customer questions
Per location, in the language people search in there, and in plain words: “best pasta near me”, not “premium Italian dining”.
Ask each one in each assistant you care about
From the location’s own city, and log the date, time, assistant, question and place you asked from.
Record two things every time
Were you named at all, and were you named first? They move separately, so one combined score hides what’s happening.
Repeat weekly for at least four weeks before drawing conclusions
One run tells you about that run; four weekly runs give you a first baseline to compare later weeks against.
Keep your published facts the same everywhere
Name, address, phone, hours, category and menu should match wherever you publish them; that part is in your hands, whatever any assistant does. Neither study shows what assistants read.
Compare against yourself
After any change, compare your new weeks with your old weeks, not with a provider’s average.
Keep reading
Making the list is only the first step
Appearing on an AI assistant’s short list and being the one picked from it are two different events.
- Making the list is not the same as being chosenWhat studies of AI recommendations show about who gets named, who comes first, and why neither is a customer yet.Read the Briefing
- AI visibility for local businessesHow Obenan helps local businesses check what AI assistants say about them.See how it works
Sources
Where this comes from
We wrote this from two studies, each published by the provider that ran it. We didn’t rerun them, and the trattoria and its answers on this page are invented.
- Four AIs, one question, four different answers
Yext Checked
The 491,972 shared local cases and the 4% four-engine agreement on the top business. Dated September 2026 with no day; study window February–August 2026.
- What 200k local AI searches tell us about what businesses get recommended, and why your website matters
BrightLocal Published Checked
The 200,085 localized searches, the 20–33% average overlap between repeat runs and the roughly 50% persistence.