GEO courseModule 5 of 10: Local and ReviewsLesson 5.2
The Reviews AI Reads: Volume, Recency, Replies and Themes
AI tools read your reviews and compress them into a line or two: "customers praise communication, but some mention slow repairs". This lesson covers the GEO angle: which review sites AI uses, what its summaries are likely to say, and how to change them honestly. Collecting and replying to reviews step by step is in our free Reviews course.
Why it matters
AI summarises what your reviews say, good and bad. A recurring complaint becomes the headline, and a profile with no recent reviews gives it little to say at all.
The evidence
Evidence: Practitioner view
That AI summarises review themes is observed but not documented. The rules on collecting reviews are official (Google policy and UK law).
Do this first
Ask three assistants "What do customers say about [your agency] in [town]?" and compare their summaries with your actual reviews.
01What it is
If twenty classmates all say the teacher is fair but the lessons are long, a new pupil will hear “fair but long”. The summary comes from what lots of people actually said, so the way to change it is to change the lessons.
Think of it like
A school report. The teacher's summary comes from many marks across the term, so improving the work changes it more than arguing with it does.
02Why it matters
To the people searching
Buyers and tenants read recent reviews about viewings, communication and fees. Sellers and landlords read about results, honesty and how problems were handled. Now they often read an AI's summary of all of that before they read a single review.
buyers
See themes about viewings, responsiveness and honesty.
tenants
See themes about repairs, deposits and communication.
sellers
See themes about results, valuations and sales progression.
landlords
See themes about rent collection, tenant finding and management.
To Google, Bing and AI
Evidence: Practitioner viewGoogle's AI features lean on Business Profile data, which includes reviews, and Ask Maps draws on them too. Other assistants cite review sites when they find them. In one UK snapshot of Google AI Mode answers about local businesses (Whito, August 2026), most answers cited Google reviews or Business Profiles. Nobody publishes a minimum number of reviews, or how summaries are made. The rules are clearer: Google bans paid and incentivised reviews, staff reviews and asking only happy customers, and UK law (the DMCC Act, since April 2025) bans fake reviews, hidden incentives and cherry-picking which reviews you show.
An agent example
A lettings agency exports its last 80 reviews: 22 mention friendly staff, 9 mention slow maintenance updates, and 6 landlords praise rent collection. AI answers describe it as "friendly, but maintenance updates can be slow". The agency introduces a weekly maintenance update email, replies to the old reviews saying what changed, and six months later new reviews start mentioning the updates.
03Your turn
Check yours, then fix it
- 1In a private window, ask ChatGPT, Gemini, Copilot, Perplexity and Claude: "What do customers say about [agency] in [town]?"
- 2Then: "What are common complaints about [agency]?"
- 3Record the themes and the sources cited. Check each complaint: is it real, old, or about a different firm?
- 1Copy your last 50 to 100 reviews from Google and one other platform, with names removed.
- 2Group them into themes: communication, speed, fees, honesty, area knowledge, viewings, repairs.
- 3Tag each positive or negative, and by audience. Look for points that repeat, not one-offs.
- 1Count your reviews, average rating, date of the latest review, how many have a reply, and how many came in the last six months.
- 2Do the same for your top three competitors.
- 3Open your review link on a phone and check it works.
Here are 50 reviews of my estate agency with all names removed. Summarise the recurring themes by audience (buyer, tenant, seller, landlord). Which are praised? Which complaints repeat? Quote the review text behind each theme. Flag any review that might break Google's review policy, without accusing anyone.
- 1Add a review request to your completion and move-in routine, for every client.
- 2Ask open questions ("What worked well? What could we have done better?"). Never supply wording or keywords.
- 3Use a QR code for your review link at sign-up and handover.
- 4No incentives, no prize draws, no staff reviews.
- 1Pick the top two complaints from your theme table each quarter.
- 2Give each an owner and change the process behind it.
- 3Reply to past reviews on that theme with what changed, factually.
- 1Reply within a few working days, calmly and specifically.
- 2For negative reviews: apologise where right, give facts, offer an offline route, and reveal no personal details.
- 3Report reviews that break the platform's policy (fake, conflict of interest, off-topic) through its own route, and keep screenshots.
- 4Never respond to an attack by posting or buying positive reviews.
- 1Pick one your clients already use (Trustpilot, AllAgents or similar), claim it and invite every client.
- 2Read its rules on invitations and display before you embed anything.
- 3On your own site, show a fair overall picture with a link to the full profile. Never show only the five-star ones and call it your rating.
Draft a calm, specific reply to this review for an estate agency. Use only these facts about what happened: [facts]. Offer to talk offline. Don't reveal personal details or admit anything I haven't said.
Then: Re-run the "what do customers say" prompts each quarter. Summaries change slowly, so look for the new theme appearing in new reviews first.
04Keep it up
Reply weekly. Read your themes quarterly and pick the next two service fixes.
That's lesson 5.2
Worked through it? Mark it done and we’ll keep track on this device.
← Back to 5.1 Google Business Profile, and Why the Map and AI Answers Differ