{"course":{"slug":"ai-for-small-business","title":"AI for a Small Business","url":"https://addaly.com/learn/ai-for-small-business","level":"beginner","track":"work"},"lesson":{"slug":"reviews-and-reputation","title":"Reviews: asking, answering, and the line you must not cross","url":"https://addaly.com/learn/ai-for-small-business/reviews-and-reputation","position":35,"of":88,"minutes":9,"answer":"Fabricating reviews is prohibited in a growing number of regimes and detectable by platforms, asking only your happy customers is gating and is banned on major platforms.","keyIdea":"Fabricating reviews is prohibited in a growing number of regimes and detectable by platforms, asking only your happy customers is gating and is banned on major platforms, and the highest-value use of AI here is reading two years of reviews in bulk to find what customers value that you never advertise."},"body":{"format":"markdown","text":"## The one absolute rule\n\nDo not fabricate reviews. Not one, not to balance an unfair one, not written by AI in a customer's voice with their permission.\n\nThis is not a matter of taste. In the United States a rule specifically prohibiting fake reviews and testimonials took effect in 2024, with penalties per violation. Consumer protection and unfair-trading rules across the EU, UK, India, Australia and elsewhere reach the same conclusion by different routes. Platforms detect patterns (timing, phrasing, device, reviewer history) and the sanction is usually removal of your listing rather than a fine.\n\nThe commercial reason is simpler than the legal one. Being caught faking reviews is a story that outlives the business, and it is discoverable by anyone with a browser.\n\n## Asking is fine. Steering is not.\n\nThe distinction matters and it is where well-meaning owners go wrong.\n\n**Fine nearly everywhere:** asking every customer for a review, at a good moment, with a direct link.\n\n**Risky:** offering a discount or entry to a prize draw in exchange. Some platforms ban it outright, some jurisdictions treat it as an inducement that must be disclosed, and the review itself must disclose it in several regimes.\n\n**Not fine:** asking only the customers you expect to be happy. This is called review gating, it is explicitly prohibited by several platforms, and it is a practice sold to small businesses as reputation management.\n\nAsk everybody, at the same moment in the process, every time. That is both the safe position and, over a year, the more useful one, because a rating built only from your happiest customers tells you nothing you can act on.\n\n## What AI helps with\n\n**The ask.** A short message that does not grovel, in the customer's language, sent at the right moment. Draft it once, use it forever; this is a template job.\n\n**The reply.** Replying to reviews is worth doing and nobody has time. Drafting is a good fit: same shape, varying content, easy to check.\n\n**Reading them in bulk.** This is the underused one. Paste two years of reviews and ask what people praise, what they complain about, what has changed over time, and what they mention that you never mention in your own marketing.\n\nThat last question earns its place regularly. Customers routinely value something the business has never advertised: the parking, the fact that you answer the phone, that you explain things without condescension. That is free market research from people who paid you.\n\n## Replying, briefly\n\nThree rules that survive every platform.\n\n**Short.** Under sixty words. Long replies read as defensive regardless of content, and a model asked to be thorough writes long.\n\n**No new facts.** Do not litigate the account in public. One fact if it is needed, then off the platform.\n\n**Never argue.** You are not writing to the reviewer. You are writing to the next hundred people who read it, and they are judging your composure rather than the merits.\n\nFor a bad review that is unfair, the reply that works is bland: acknowledge, state you would like to sort it out, give a direct contact. Readers score that as reasonable. A point-by-point rebuttal, however correct, reads as the business being difficult.\n\n## Disclosure, where it applies\n\nIf you use AI to draft your replies, you do not need to announce it. A reply is your statement whoever typed it, and nobody expects otherwise.\n\nTwo things do need care. Do not present AI-generated text as somebody else's words: a customer testimonial, a staff quote. And where a platform or a jurisdiction requires synthetic media to be labelled, that includes images and video in your marketing, which module seven covers.\n\n## The one that is a business fix\n\nIf the same complaint appears in three reviews, no reply strategy helps. Fix the thing.\n\nBulk analysis is useful precisely because it makes that pattern visible, and because it is hard to argue with when it is your own customers saying it in their own words.\n\n## When to ask\n\nTiming matters more than wording, and it is free to get right.\n\nAsk when the customer has just experienced the good part: at collection, on delivery, after the job is signed off, the morning after the stay. Not a week later, when the feeling has faded and the message reads as marketing.\n\nFor a service with a long tail (a repair, a treatment, a build) ask twice at most, and stop. Repeated asking annoys the people most likely to have written something good.\n\n## Keep the raw text\n\nWhatever platform your reviews live on, keep your own copy. Export them quarterly into a spreadsheet, or paste them into a document.\n\nPlatforms remove reviews, change their display, and occasionally lose them. More practically, the bulk analysis in this lesson works on text you hold, and it is far easier to ask two years of questions of a file you own than of a website that shows ten at a time."},"check":{"question":"A restaurant sends its review request only to diners who told the server they enjoyed the meal. The owner sees this as sensible targeting. What is the problem?","options":[{"text":"Selective requesting is inefficient, because satisfied diners are the least likely to take the time to write a review without an incentive.","correct":false,"tempting":"Argues from response rates rather than from the rule, and would be answered by adding an incentive, which introduces a separate disclosure problem."},{"text":"The resulting rating misrepresents the customer base, which is a marketing weakness rather than a compliance issue for a business of this size.","correct":false,"tempting":"Correctly spots the distortion and then downgrades it to a matter of taste, when several major platforms prohibit the practice outright regardless of business size."},{"text":"This is review gating (filtering who is asked so the rating reflects only expected positives) which major platforms prohibit and which also destroys the feedback's usefulness.","correct":true},{"text":"Asking diners in person creates social pressure that may amount to an inducement under consumer protection rules in some jurisdictions, particularly where the server is present when the review is written on a device belonging to the restaurant.","correct":false,"tempting":"Constructs an inducement argument out of ordinary conversation, and misses that the defect is in who gets asked rather than in how the asking happens."}]},"prev":{"slug":"bookings-reminders-no-shows","title":"Bookings and no-shows: the case where AI is the wrong tool","url":"https://addaly.com/learn/ai-for-small-business/bookings-reminders-no-shows"},"next":{"slug":"phone-and-voice","title":"The phone: what a voice assistant can and cannot do yet","url":"https://addaly.com/learn/ai-for-small-business/phone-and-voice"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Reviews: asking, answering, and the line you must not cross', in AI for a Small Business. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-small-business/reviews-and-reputation (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Reviews: asking, answering, and the line you must not cross","courseTitle":"AI for a Small Business","module":{"index":4,"slug":"customers-and-the-front-desk","title":"Customers, from first message to resolved complaint"},"answer":null,"keyIdea":"Fabricating reviews is prohibited in a growing number of regimes and detectable by platforms, asking only your happy customers is gating and is banned on major platforms, and the highest-value use of AI here is reading two years of reviews in bulk to find what customers value that you never advertise.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-one-absolute-rule","runs":[{"text":"The one absolute rule"}]},{"t":"p","runs":[{"text":"Do not fabricate reviews. Not one, not to balance an unfair one, not written by AI in a customer’s voice with their permission."}]},{"t":"p","runs":[{"text":"This is not a matter of taste. In the United States a rule specifically prohibiting fake reviews and testimonials took effect in 2024, with penalties per violation. Consumer protection and unfair-trading rules across the EU, UK, India, Australia and elsewhere reach the same conclusion by different routes. Platforms detect patterns (timing, phrasing, device, reviewer history) and the sanction is usually removal of your listing rather than a fine."}]},{"t":"p","runs":[{"text":"The commercial reason is simpler than the legal one. Being caught faking reviews is a story that outlives the business, and it is discoverable by anyone with a browser."}]},{"t":"h2","id":"asking-is-fine-steering-is-not","runs":[{"text":"Asking is fine. Steering is not."}]},{"t":"p","runs":[{"text":"The distinction matters and it is where well-meaning owners go wrong."}]},{"t":"p","runs":[{"text":"Fine nearly everywhere:","b":true},{"text":" asking every customer for a review, at a good moment, with a direct link."}]},{"t":"p","runs":[{"text":"Risky:","b":true},{"text":" offering a discount or entry to a prize draw in exchange. Some platforms ban it outright, some jurisdictions treat it as an inducement that must be disclosed, and the review itself must disclose it in several regimes."}]},{"t":"p","runs":[{"text":"Not fine:","b":true},{"text":" asking only the customers you expect to be happy. This is called review gating, it is explicitly prohibited by several platforms, and it is a practice sold to small businesses as reputation management."}]},{"t":"p","runs":[{"text":"Ask everybody, at the same moment in the process, every time. That is both the safe position and, over a year, the more useful one, because a rating built only from your happiest customers tells you nothing you can act on."}]},{"t":"h2","id":"what-ai-helps-with","runs":[{"text":"What AI helps with"}]},{"t":"p","runs":[{"text":"The ask.","b":true},{"text":" A short message that does not grovel, in the customer’s language, sent at the right moment. Draft it once, use it forever; this is a template job."}]},{"t":"p","runs":[{"text":"The reply.","b":true},{"text":" Replying to reviews is worth doing and nobody has time. Drafting is a good fit: same shape, varying content, easy to check."}]},{"t":"p","runs":[{"text":"Reading them in bulk.","b":true},{"text":" This is the underused one. Paste two years of reviews and ask what people praise, what they complain about, what has changed over time, and what they mention that you never mention in your own marketing."}]},{"t":"p","runs":[{"text":"That last question earns its place regularly. Customers routinely value something the business has never advertised: the parking, the fact that you answer the phone, that you explain things without condescension. That is free market research from people who paid you."}]},{"t":"h2","id":"replying-briefly","runs":[{"text":"Replying, briefly"}]},{"t":"p","runs":[{"text":"Three rules that survive every platform."}]},{"t":"p","runs":[{"text":"Short.","b":true},{"text":" Under sixty words. Long replies read as defensive regardless of content, and a model asked to be thorough writes long."}]},{"t":"p","runs":[{"text":"No new facts.","b":true},{"text":" Do not litigate the account in public. One fact if it is needed, then off the platform."}]},{"t":"p","runs":[{"text":"Never argue.","b":true},{"text":" You are not writing to the reviewer. You are writing to the next hundred people who read it, and they are judging your composure rather than the merits."}]},{"t":"p","runs":[{"text":"For a bad review that is unfair, the reply that works is bland: acknowledge, state you would like to sort it out, give a direct contact. Readers score that as reasonable. A point-by-point rebuttal, however correct, reads as the business being difficult."}]},{"t":"h2","id":"disclosure-where-it-applies","runs":[{"text":"Disclosure, where it applies"}]},{"t":"p","runs":[{"text":"If you use AI to draft your replies, you do not need to announce it. A reply is your statement whoever typed it, and nobody expects otherwise."}]},{"t":"p","runs":[{"text":"Two things do need care. Do not present AI-generated text as somebody else’s words: a customer testimonial, a staff quote. And where a platform or a jurisdiction requires synthetic media to be labelled, that includes images and video in your marketing, which module seven covers."}]},{"t":"h2","id":"the-one-that-is-a-business-fix","runs":[{"text":"The one that is a business fix"}]},{"t":"p","runs":[{"text":"If the same complaint appears in three reviews, no reply strategy helps. Fix the thing."}]},{"t":"p","runs":[{"text":"Bulk analysis is useful precisely because it makes that pattern visible, and because it is hard to argue with when it is your own customers saying it in their own words."}]},{"t":"h2","id":"when-to-ask","runs":[{"text":"When to ask"}]},{"t":"p","runs":[{"text":"Timing matters more than wording, and it is free to get right."}]},{"t":"p","runs":[{"text":"Ask when the customer has just experienced the good part: at collection, on delivery, after the job is signed off, the morning after the stay. Not a week later, when the feeling has faded and the message reads as marketing."}]},{"t":"p","runs":[{"text":"For a service with a long tail (a repair, a treatment, a build) ask twice at most, and stop. Repeated asking annoys the people most likely to have written something good."}]},{"t":"h2","id":"keep-the-raw-text","runs":[{"text":"Keep the raw text"}]},{"t":"p","runs":[{"text":"Whatever platform your reviews live on, keep your own copy. Export them quarterly into a spreadsheet, or paste them into a document."}]},{"t":"p","runs":[{"text":"Platforms remove reviews, change their display, and occasionally lose them. More practically, the bulk analysis in this lesson works on text you hold, and it is far easier to ask two years of questions of a file you own than of a website that shows ten at a time."}]}],"check":{"question":[{"text":"A restaurant sends its review request only to diners who told the server they enjoyed the meal. The owner sees this as sensible targeting. What is the problem?"}],"options":[{"text":[{"text":"Selective requesting is inefficient, because satisfied diners are the least likely to take the time to write a review without an incentive."}],"correct":false,"tempting":[{"text":"Argues from response rates rather than from the rule, and would be answered by adding an incentive, which introduces a separate disclosure problem."}]},{"text":[{"text":"The resulting rating misrepresents the customer base, which is a marketing weakness rather than a compliance issue for a business of this size."}],"correct":false,"tempting":[{"text":"Correctly spots the distortion and then downgrades it to a matter of taste, when several major platforms prohibit the practice outright regardless of business size."}]},{"text":[{"text":"This is review gating (filtering who is asked so the rating reflects only expected positives) which major platforms prohibit and which also destroys the feedback’s usefulness."}],"correct":true,"tempting":null},{"text":[{"text":"Asking diners in person creates social pressure that may amount to an inducement under consumer protection rules in some jurisdictions, particularly where the server is present when the review is written on a device belonging to the restaurant."}],"correct":false,"tempting":[{"text":"Constructs an inducement argument out of ordinary conversation, and misses that the defect is in who gets asked rather than in how the asking happens."}]}],"order":[1,3,0,2],"why":null},"prev":{"slug":"bookings-reminders-no-shows","title":"Bookings and no-shows: the case where AI is the wrong tool","minutes":9},"next":{"slug":"phone-and-voice","title":"The phone: what a voice assistant can and cannot do yet","minutes":9}}}