Local Visibility
Reviews and AI: What Buyers Check After AI Names You
Published July 27, 2026
Reviews are not the input to the AI answer. They are what the buyer checks after the answer names you. We have found no study showing a review profile changes the answer, and in BrightLocal's Local Consumer Review Survey 2026, 88% of people who used AI for a local recommendation fact-check what they are told. So there are two ways to lose the job: the assistant never names you, or your reviews talk the buyer out of it.
What do the studies about reviews and AI answers actually show?
Less than the usual claim, that a strong review profile is why an assistant names a business. The studies offered for it measure something narrower: how often review platforms appear as cited sources. Being cited is not being the reason.
The strongest study states its scope up front. SE Ranking checked 30,000 US commercial keywords in December 2025: 34.5% of the AI Overviews cited at least one review platform, but those were business software queries, where G2, Capterra, and Gartner Peer Insights dominate and Yelp sat under 1%. Two-thirds cited none, and none of it transfers to a plumber in Fishers. The sources local answers actually quote get an article of their own, and that research counts citations too, not causes. SE Ranking and BrightLocal both sell tools in this market.
What does a buyer do after an assistant names a business?
They check. That 88% comes from the 455 respondents in BrightLocal's Local Consumer Review Survey 2026 who had used an AI tool for a local business recommendation in the last 12 months: 51% of them check whether the reviews are legitimate, and 37% check the source. Across the same group, 97% at least sometimes double-check against real reviews.
In BrightLocal's Consumer Search Behavior study, published July 2026, only 18% of consumers already using AI for local search felt ready to contact the business they were given. The rest kept researching: an AI mention is not a lead. Getting named at all is a separate job. Your reviews are part of what they find while they keep looking.
Where do buyers go to check?
On Google, more often than on social media. Of the consumers in that July 2026 study who started on an AI tool, 43% went on to Google, 39% to social media, and 2% used the assistant as their only channel. AI has not replaced Google.
- Google Business Profile first. The reviews and the listing sit in the same panel, and the listing around them is a job of its own.
- Then whatever comes up when you search your service plus your city. Include your trade's own platform if it has one.
- Depth over breadth. Two surfaces with recent, specific reviews satisfy the check better than a thin scattering across ten, and a claimed but empty profile reads as shut. A perfect 5.0 is not the goal either: a few imperfect reviews with calm replies read as real, and a wall of identical five-star one-liners does not.
How do you run the ask without breaking a rule?
One ask moment, one link, one reminder. Five moments is the same as none: anything that depends on remembering does not survive a busy week.
- Pick the one moment and give it one owner. For home services it is job completion, while the technician is still on site or within the hour: a link the customer opens on their own time, never a review written while you wait.
- Put the ask where the customer already is. The invoice, the job-completion text, the receipt. One link straight to the review form, plus a QR code where the work is physical.
- Send it from the person who did the work. The technician who was just in the house, not an unfamiliar address.
- One reminder at three to five days, then stop. A sequence reads as pressure.
- Configure it inside the tool you already run. A routine that needs new software stalls.
The rails are not preferences. Google's Maps content policy prohibits incentivized reviews and review gating, and Google enforces it. The FTC's rule on consumer reviews and testimonials (16 CFR Part 465) bans fake reviews, buying reviews, and suppressing negative ones.
- Nothing offered for a review. Nothing offered to get a negative one changed either. That includes staff: no leaderboards and no per-review bonuses.
- No gating. A pre-survey that sends only happy customers to the public link is review gating.
- No dictating content. Never ask for a service, a keyword, a city, or a staff member to be named, and never write or post a review as the customer.
- No Yelp asks. Yelp prohibits soliciting reviews, so Yelp work is profile hygiene only.
- Healthcare adds a step. The practice's own licensed clinician signs off on the routine in writing, and the ask never references a treatment, a condition, or the reason for the visit.
What is the reply for?
The next buyer, not the reviewer. The person deciding whether to call arrived from an answer, went looking for the worst reviews, and is reading the exchange. Reply to every review inside two business days, in the business's own name.
- Calm and specific. Name what happened in the customer's terms, without a paragraph of defense.
- Own your part and state the fix. "We missed the window twice, and we have changed how we confirm same-day appointments" does more than an apology.
- Never argue, and never disclose. Appointment details, anything clinical, and anything the customer did not publish stay out of a public reply.
- Keep the good ones short, and keep going. Replies carry dates, and a last response two years old tells the buyer more than the rating does.
Which numbers tell you it is working?
Five, and the first three have to be captured before anything changes. In the week you start, record the review count, the rating, and the trailing 90 days of reviews divided by three: your monthly velocity before the work. Without that baseline, every later reading is an opinion.
- Count and rating. The obvious pair, read monthly rather than weekly.
- Velocity against the baseline. Reviews per month now against the pre-work figure: the number that says whether the routine is running.
- Response rate. Reviews carrying a reply divided by reviews received, plus the age of the oldest unanswered one.
- Ask rate. Asks sent divided by completed jobs, reported by you from your own system. Flat velocity with a low ask rate is a workflow problem, not a copy problem.
The free Visibility Check scores five categories and marks the ones that are estimates. Its Review Presence category is a signal-based estimate: it checks whether a Google Business Profile link was provided and which review platforms your site links to, not your reviews or your counts. The routine itself is part of the AI visibility program: Foundations puts the review link and response routine in place, and Growth and up builds the cadence into how you already close a job. Book a demo to walk through the ask and the replies with someone.
Sources
- BrightLocal, "Local Consumer Review Survey 2026," March 2026 (brightlocal.com) · The survey of 1,002 US adults behind the AI figures used here, with the AI questions answered by the 455 respondents who had used an AI tool for a local business recommendation in the previous 12 months. BrightLocal sells local marketing software.
- BrightLocal, "Local Consumer Review Survey 2026" (AI focus report), March 2026 (brightlocal.com) · Among those 455 AI users: 88% fact-check what they are told, 51% check specifically whether the reviews are legitimate, 37% check the source, and 97% at least sometimes double-check against real reviews.
- BrightLocal, "Consumer Search Behavior 2026," July 2026 (brightlocal.com) · The survey of 1,227 US consumers who had searched for a local business in the previous three months: 18% of those already using AI for local search felt ready to contact the recommended business; of those who started on an AI tool, 2% used it as their only channel, 43% went on to check Google, and 39% checked social media.
- SE Ranking, "Review platforms in AI Overviews," January 2026 (seranking.com) · 34.5% of AI Overviews citing at least one review platform and two-thirds citing none, across 30,000 commercial keywords in the US on December 2025 data; the platforms being G2, Capterra, and Gartner Peer Insights, with Yelp under 1%. Business software queries, not local services. SE Ranking sells SEO software.
- Google, "Prohibited and restricted content" (support.google.com) · Google's user contributed content policy for Maps prohibiting incentivized reviews and review gating, as it read when we checked in September 2026.
- U.S. Government Publishing Office, "Code of Federal Regulations, Title 16 Part 465," 2025 (govinfo.gov) · 16 CFR Part 465, the FTC's rule on consumer reviews and testimonials, in force: it bans fake reviews, buying reviews, and suppressing negative reviews. Cited for the existence and substance of the rule only.
- Yelp, review solicitation policy (yelp.com) · Yelp asks businesses not to solicit reviews, which is why no Yelp ask appears in this routine. Cited for the existence of the policy; no link in this entry.
Questions people ask
We cannot show that, so we do not claim it. The available studies count how often review platforms appear as sources in AI answers, which is a different thing from one business's profile changing the answer. SE Ranking's December 2025 check of 30,000 commercial keywords found 34.5% of the AI Overviews citing a review platform, but those were business software keywords where G2 and Capterra dominate and Yelp was under 1%. The measured part is the buyer's check afterward.
There is no number to hit. The bar that matters is the businesses a buyer sees beside you when they check: a local comparison, not a threshold. Build the ask into one moment you run on every job, and measure against your own baseline from before you started. Treat any count a vendor quotes as a sales figure until they show you the study.
Only by fixing the problem, and with nothing offered in exchange. Resolve it, tell them what changed, and leave the decision with them. Tying a refund or a discount to a review edit is an incentive, which Google's content policy prohibits, and the FTC rule on consumer reviews bans buying reviews and suppressing negative ones. Reply calmly in public instead: that is what the verifying buyer reads.
No. Yelp prohibits soliciting reviews, so we never ask for Yelp reviews, for ourselves or for a client. Claim the page, keep the details accurate, and reply to whatever arrives there. The ask itself goes to Google, and to your trade's own platform if it permits one.