It is Sunday evening and your Google Business Profile shows five new reviews. Three are short five-star thank-yous. One says the food was good but the wait was long. One gives a single star, says the customer's son reacted to a sauce and claims staff ignored the allergy question. A review tool you are trialling offers to answer all five in one click. Should you let it?
Short answer: you can use AI to respond to Google reviews, but not to all of them in the same way. AI is useful for drafting short replies to routine positive reviews, spotting themes and making sure nothing goes unanswered. A person should check every reply that states facts, admits or denies fault, mentions health or safety, offers money, or answers a review you suspect is fake. UK law and Google's own policies are strict about the reviews themselves: no fake reviews, no hidden incentives, no burying negative ones. AI-drafted replies are not banned, but a reply that misleads customers is still your business's statement. Below we cover what the rules say, what AI does well, where it goes wrong, how the time saving looks on a hypothetical example, and a checklist for deciding. This is general information, not legal advice.
Reviews and replies are covered by different rules
The strictest UK rules apply to reviews, not replies. Since 6 April 2025, paragraph 13 of Schedule 20 to the Digital Markets, Competition and Consumers Act 2024 (the DMCC Act) lists these as banned practices in all circumstances:
- submitting, or commissioning someone else to submit or write, a fake consumer review or one that hides that it was incentivised;
- publishing reviews or review information "in a misleading way", which includes removing or not publishing negative reviews while publishing positive ones, or giving positive reviews greater prominence;
- publishing reviews without "reasonable and proportionate steps" to prevent and remove fake reviews, reviews that hide an incentive and false or misleading review information.
Two definitions matter for AI. A fake review is one that "purports to be, but is not, based on a person's genuine experience", and "writing" a review "includes creating by any means". A review that an AI tool generates for your business is a fake review, however natural it sounds.
The Competition and Markets Authority (CMA) can fine a business directly for breaking consumer law. Under section 182 of the DMCC Act, a penalty can be up to £300,000 or, if higher, 10% of the business's turnover.
The CMA's fake reviews guidance (CMA208) gives examples of what it treats as commissioning a banned review. They include "contacting a customer who has left a negative review and offering them a refund and/or a gift card if they change their review to remove the negative commentary", and "buying reviews which look like they have been written by individual consumers but have in fact been generated by software applications (such as bots)". The guidance also says a review "is not fake just because the trader who it is about does not like or agree with it, as long as it reflects the reviewer's genuine experience". Its short guide for businesses covers the duties of traders who publish reviews, for example on their own website.
The rules are being enforced. In a CMA blog post of 14 October 2026, the regulator said it continues to progress investigations into Autotrader, Feefo, Dignity, Just Eat and Pasta Evangelists over concerns that the law on misleading reviews may have been broken, with a further update expected in the coming months. These are investigations, not findings. The same post warns that commissioning reviews "isn't limited to paying consumers for them", and that unlawful conduct could include "a business requesting members of staff to write reviews about its products or services". On AI generally it adds: "New technology does not change their responsibility to comply with consumer law."
Replies are different. Nothing in paragraph 13 bans using AI to draft them. But a reply is your business speaking to potential customers. Section 226 of the DMCC Act treats the provision of "false or misleading information" about a product or trader as a misleading action. In our reading, an AI-drafted reply that wrongly says a dish is nut-free, that a problem has been fixed, or that a refund was issued could fall under this, in the same way as if a manager had typed it.
What Google allows on Business Profiles
Google's rules sit on top of the law. Its prohibited and restricted content policy for Maps says reviews must reflect "a genuine experience". Merchants may not offer incentives for posting a review or for changing or removing a negative one, and may not "discourage or prohibit negative reviews, or selectively solicit positive reviews from customers". The policy also lists reviews based on a conflict of interest, including current or former employment, and asking staff to collect reviews that mention a particular staff member.
If Google decides a business has broken its fake engagement policy, its page on Business Profile restrictions says it may stop the profile receiving new reviews for a set period, unpublish existing reviews for a set period, or show customers a warning that fake reviews were removed.
For replies, Google's help page on managing reviews says replies are checked against its content policies before they appear. That usually takes up to 10 minutes but can take up to 30 days. Approved replies are public, appear as the business rather than a named person, and the reviewer is notified and can still change their review. Google's tips for review replies are worth reading before you configure any tool:
- "Respond if necessary": reply when you have new or relevant information, because "you don't need to thank every reviewer publicly".
- "Be a friend, not a salesperson": no deals or promotions in replies.
- For negative reviews: never share the reviewer's private information, check your records about the visit, admit mistakes but explain what is outside your control, and sign the reply with your name or initials.
Where AI helps with review replies
Most of the time spent on reviews is reading, sorting and writing the first draft. That is where AI tools are useful:
- Drafting routine thank-yous. A short positive review that names a dish, a job or a member of staff can get a short, specific reply drafted in your house style. You decide whether it goes out.
- Sorting and flagging. AI can label each review as positive, mixed or negative, pick out topics such as waiting times, prices or cleanliness, and flag words that need a person: allergy, injury, refund, solicitor, discrimination.
- Spotting patterns. A monthly summary of what customers keep mentioning is often more valuable than any single reply.
- Making sure nothing is missed. A list of unanswered negative reviews is a simple and reliable job for software.
Google offers some of this itself. Its help page on managing a Business Profile in the Gemini web app says Gemini can summarise reviews and "draft or delete replies". Google lists conditions: you must manage only one verified profile, be 18 or over and sign in with a personal Google Account rather than a work account. Third-party review tools can connect through Google's Business Profile API, which lets software list reviews and post or delete replies. Whether replies post automatically or wait for approval depends on the tool and its settings, so check this before you connect your profile.
What to keep human
AI language models write fluent text, which is the problem here: a confident wrong reply reads as well as a right one. Keep a person on these:
- Any reply that states facts about the visit or job. The model does not know what happened. Google's own advice is to check your records first. A draft that says "we have spoken to the team" when nobody has is a false statement.
- Health, safety and allergens. Never let AI say what a dish contains, whether a mistake happened or what you have changed. These replies need someone who knows the facts, and possibly your insurer or an adviser.
- Offers of money or goodwill. You can invite an unhappy customer to contact you and you can refund them. What you cannot do is make a refund or voucher conditional on changing or removing the review: both CMA208 and Google's policy name this as banned. Do not let AI improvise offers.
- Admitting or denying fault. An apology is often right; an admission of something that did not happen, or a denial of something that did, is a problem either way.
- Suspected fake reviews and extortion. In its guidance on review extortion scams, Google describes a sudden wave of 1 and 2-star reviews followed by a demand for money, and says: "Do not engage with or pay the malicious individuals." Report these to Google rather than letting a tool reply to each one.
- Anything personal. Booking details, health information, staff disciplinary matters and the reviewer's identity stay out of public replies.
A simple split: automate, draft or handle personally
| Review type | AI role | Who approves | Why |
|---|---|---|---|
| Star rating only, no text | None, or a short generic thanks | Optional | Google says you do not need to thank every reviewer publicly |
| Short positive review, no claims | Draft a short, specific reply | Quick check, or automatic once you trust the tone | Low risk; the draft only thanks and refers to what the customer said |
| Mixed review with a specific point | Draft, flag the point raised | Manager | The reply should say what is true about the point, which AI cannot know |
| Negative review about service or price | Summarise, suggest a draft | Manager or owner | Needs a check of records and a decision on what to offer, if anything |
| Health, safety, allergens, injury, legal threat | Flag only | Owner, with advice if needed | Facts, liability and possibly insurance are involved |
| Suspected fake or extortion wave | Flag only | The owner | Report to Google; do not engage or offer anything |
Review tool features to switch off or avoid
Some review software sold to small businesses goes beyond replies. When you trial a tool, check for these features:
- Review generation. Any feature that writes reviews for you or your staff to post creates fake reviews under the DMCC Act definition. Drafting reviews for customers to post steers what they say, which Google's policy does not allow and which can make the review fake.
- Review gating. Some tools first ask customers how their visit went, then send only satisfied customers to Google and route unhappy ones to a private form. Google's policy bans selectively soliciting positive reviews, and CMA208 describes "encouraging just those who are satisfied to leave reviews" as a form of cherry-picking.
- Incentives in replies or requests. Discount codes for reviewers, or replies offering money to "update" a review.
- Staff review targets. Google's policy lists asking staff to collect a certain number of reviews, or reviews that mention a staff member, as not allowed.
- Fully automatic posting for every review. Convenient, but it removes the human check exactly where it matters most.
Asking every customer for a review is fine. CMA208 says emailing customers generally to ask whether they would like to leave a review, without steering what they say, is not prohibited, and Google's tips suggest sharing a review link or QR code.
A worked example: how much time AI drafting saves
This example is hypothetical. The figures are our assumptions, not measurements from any business or tool. A café gets 30 Google reviews a month: 24 short positive ones, 4 mixed and 2 negative. The owner values their time at £25 an hour. Writing a reply by hand takes about 3 minutes for a positive review and 15 minutes for a mixed or negative one, including checking what happened. With AI drafts, checking a positive reply takes about 1 minute; a mixed or negative one still takes about 10, because checking the facts does not get faster.
| Replies | By hand | With AI drafts |
|---|---|---|
| 24 positive replies | 24 × 3 min = 72 min | 24 × 1 min = 24 min |
| 6 mixed or negative replies | 6 × 15 min = 90 min | 6 × 10 min = 60 min |
| Total | 162 min (2.7 h), about £67.50 | 84 min (1.4 h), about £35 |
On these assumptions AI drafting saves 78 minutes, about £32.50 of the owner's time, a month. A tool that costs more than that does not pay for itself in time alone; its value would have to come from consistency, faster replies and not missing a bad review. Google notes that you don't need to thank every reviewer publicly, so replying only where you have something to add would cut the by-hand time too. Most of the saving is on routine positive reviews. The negative ones still need the owner, and they are the replies future customers read most closely.
A checklist before you switch it on
AI drafting for review replies is more likely to be worth it if:
- you get enough reviews that replies regularly fall behind, or run several locations;
- you can write down your tone and the things replies must never say;
- someone named checks the queue at least every few days;
- the tool lets you choose which reviews post automatically and which wait for approval;
- the tool has no review generation, gating or incentive features switched on.
It is probably not worth it if you get a handful of reviews a month, if most of your reviews raise specific complaints, or if nobody will read the drafts before they go out. In that case a short set of example replies written by you is cheaper.
Whatever you choose, write down the rules: which review types AI may draft, who approves, what replies must never contain, and how you report suspected fakes. That can also form part of the "reasonable and proportionate steps" the CMA expects from businesses that publish reviews on their own website.
Getting it set up properly
The harder work is deciding which reviews post automatically, which go to a manager, what triggers an alert, and checking that the tool you connect has no features that break UK law or Google's policy.
If you want this mapped and set up for your business, Evoogen can review how you handle reviews now, set up AI drafting with approval steps and alerts for sensitive reviews, and agree a routine for checking the queue. Decisions on what you say to customers stay with you. See how Evoogen works with small businesses.
Sources & further reading
- legislation.gov.uk: Digital Markets, Competition and Consumers Act 2024, Schedule 20 (banned practices), paragraph 13 — checked 2026-11-11
- legislation.gov.uk: DMCC Act 2024, section 182 (Final infringement notice) — checked 2026-11-11
- legislation.gov.uk: DMCC Act 2024, section 226 (Misleading actions) — checked 2026-11-11
- CMA: Fake reviews guidance (CMA208), April 2025 (PDF) — checked 2026-11-11
- GOV.UK (CMA): Short guide for businesses: publishing consumer reviews and complying with consumer protection law — checked 2026-11-11
- CMA blog (Emma Cochrane), 14 October 2026: Direct consumer enforcement: 18 months on — checked 2026-11-11
- Google Maps User Contributed Content Policy: Prohibited and restricted content — checked 2026-11-11
- Google Business Profile Help: Business Profile restrictions for policy violations — checked 2026-11-11
- Google Business Profile Help: Manage customer reviews — checked 2026-11-11
- Google Business Profile Help: Tips to get more reviews (including tips to write better review replies) — checked 2026-11-11
- Google Business Profile Help: Manage your Business Profile on Google with the Gemini web app — checked 2026-11-11
- Google Business Profile Help: Report negative review extortion scams — checked 2026-11-11
- Google for Developers: Business Profile APIs, Work with review data — checked 2026-11-11
