review-ask

---
name: review-ask
description: Builds the review request you are allowed to send. It works out who must be asked rather than who you would like to ask, writes the message with no rating and no steer in it, checks any incentive against the banned practice in the Digital Markets, Competition and Consumers Act 2024 and against the platform's own rules, and logs every send so you can prove what you did. Use before you ask anybody for a review, and before you buy software that asks for you.
---

# The reviews you ask for, from everybody you served, in a way you can defend

You give this the list of people you served in a period, the platform you want the reviews on, and whether you are offering anything in return. You get back one asking message with no star rating in it, the rule that decides who is on the list, the disclosure wording if there is an incentive, the lawful route for sending it, and a log row per send. It will not build you a funnel that sends happy customers to Google and unhappy ones to a complaints inbox, because that is now a banned practice with a real fine attached.

## What it does

1. **Fix the rule that decides who is asked, and write it down before you look at a single name.** The rule is a period and a transaction type: everyone who ate in the restaurant between two dates, everyone who stayed in the hotel in a month. It is not everyone who seemed happy, not the tables the manager remembers, and not the ones who tipped. Paragraph 13(5)(i) of Schedule 20 to the Digital Markets, Competition and Consumers Act 2024 makes publishing reviews in a misleading way a banned practice, and the CMA's guidance says cherry picking "might be done either through suppressing negative reviews that have been submitted or by encouraging just those who are satisfied to leave reviews". Selection is where this goes wrong, not wording. Record the rule, the date range and the row count: that is your evidence.

2. **Delete the filter question. There is no lawful version of it.** The pattern is familiar: a text asking how did we do today, a five star button that goes to Google and a one star button that goes to a feedback form. Tripadvisor names it and bans it: "We prohibit the practice of selectively soliciting positive content and/or rejecting moderate or negative content. If any survey or external website ultimately directs users to submit a review on Tripadvisor, the user interface and experience for submitting positive and negative reviews must be identical." Google says the same in one line: merchants may not "Discourage or prohibit negative reviews, or selectively solicit positive reviews from customers". One link, one destination, everybody gets the same one.

3. **Decide the incentive question in two parts, because the law and the platform disagree and the stricter one wins.** Under the Act a review that "conceals the fact it has been incentivised" is banned, and a review is incentivised where "a person has been commissioned to submit or write the review". The CMA's guidance says traders may incentivise provided they "tell consumers that the review has been incentivised" and "the review must still reflect the reviewer's genuine experience". So the law permits a disclosed incentive. Google does not, prohibiting merchants from offering "payment, discounts, free of cost goods and/or services" for a review. Nor does Tripadvisor: "It is against our guidelines to offer or promise anything in exchange for any reviews, irrespective of rating. Examples include offers for free drinks, discounts, entry into a contest, making donations to a cause in the name of a customer etc." For a venue asking on Google or Tripadvisor the answer is no incentive at all.

4. **Take the target off your team, and take your team out of the reviews.** Google states that merchants must not "require or pressure users to leave ratings or write reviews while on the premises", and lists as prohibited "Merchants requesting that staff solicit a certain number of reviews" and "Merchants requesting that staff solicit reviews that include specific content, including content that identifies a staff member". Tripadvisor bans employee incentives outright, including "a bonus for being mentioned in a review or a contest for the highest number of reviews achieved within a certain timeframe", and bans the helpful behaviours too: "monitoring users while they write reviews", "suggesting specific terms, wording or paraphrasing to be included in reviews". A waiter asking a table to mention them by name breaks several of those at once.

5. **Nobody connected to the business writes one, and that is wider than you think.** Paragraph 13(1)(a) bans submitting or commissioning "a fake consumer review", defined as one "that purports to be, but is not, based on a person's genuine experience". On 27 March 2026 the CMA opened investigations into five businesses including, in its own words, "Whether Dignity asked staff to write positive reviews about the company's crematoria services", and said it "can fine them up to 10% of their global turnover". Tripadvisor's bias list also excludes anyone "related to or are friends with a current employee or owner", anyone who receives or provides goods or services to the property, and anyone who owns or works at a business of the same category within "16 kilometres/10 miles". Your linen supplier and the chef from the pub down the road are both barred.

6. **Work out whether your ask is direct marketing, because that decides how you may send it.** The ICO sets the test: PECR defines direct marketing as "the communication (by whatever means) of advertising or marketing material which is directed to particular individuals", and a neutral administrative message is a service message. The trap is the next line: "If your service message has elements that are direct marketing, even if that is not the main purpose of your message, then it will count as direct marketing." A plain request to write a review is arguably a service message. Add ten per cent off the next booking and the whole message is marketing, and regulation 22 of the Privacy and Electronic Communications (EC Directive) Regulations 2003 applies: consent is needed unless the details came "in the course of the sale or negotiations for the sale of a product or service", the marketing is for "similar products and services only", and a simple free refusal was offered at collection and since.

7. **Set the timing as a rule and then leave it alone.** Tripadvisor accepts a review only inside its own window: "all reviews must be submitted within one year of your experience", one review per visit, with "a minimum character limit of 25 for all reviews". An ask a fortnight after the meal sits inside every window; an ask a year later is wasted. Send once per visit, do not chase, and fix the day and the delay in writing, because the CMA lists "arbitrarily stopping and starting review invitations" among the ways a trader interferes with reviewers. A schedule that quietly pauses after a bad Saturday is exactly that.

8. **Write the message, and keep the rating out of it.** Three sentences. Say who you are and when they visited, ask them to write about their experience, give the one link. Do not name a number of stars, do not use the words positive, great, glowing or lovely, do not ask them to mention a dish, a member of staff or the parking, and do not apologise in advance. Google permits merchants to "Solicit or encourage the posting of content that does represent a genuine experience, without offering incentives to do so or attempting to influence the rating or the contents of the review". The second half of that sentence is the whole constraint, and it rules out most of what review software writes by default. Sign it with the name of the person sending it.

9. **Log the send in the form an investigator would ask for.** One row per send: date, the selection rule, the number of people on the list, the platform, the exact wording used, whether an incentive was offered and how it was disclosed, the lawful basis or PECR route, and the opt out method. If you later want to quote one of those reviews on your own website, you need more again: CAP Code rule 3.47 requires marketers to "hold documentary evidence that a testimonial or endorsement used in a marketing communication is genuine, unless it is obviously fictitious", and rule 3.50 states that marketing communications "must not feature a testimonial without permission". The log is where that evidence comes from a year later when nobody remembers.

## Then it checks

1. The selection rule is written down as a date range and a transaction type, the list was built from it, and no column in the list records how happy anyone seemed.
2. No message in the set contains a star count, a number of stars, the words positive, great, glowing, five star or lovely, or a request to mention any named dish, person or feature.
3. Every recipient receives the same single link to the same page, and no message contains a branching question, a satisfaction score, a smiley scale or any route that changes with the answer.
4. Any incentive is named in the ask, carries disclosure wording for the review itself, does not depend on the rating, and has been checked against the platform's own published policy quoted by name; where that policy prohibits it, the incentive is removed rather than reworded.
5. No message is sent by or in the name of a member of serving staff, no staff member has a review target or a bonus attached to reviews, and every name has been screened against the platform's bias rules for employees, family, friends, suppliers and nearby competitors.
6. Each send has a recorded lawful basis or PECR route, a working opt out, and a log row carrying date, rule, count, platform, exact wording, incentive status and sender name.

Any check fails: name it, redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only.
- Never invent a fact, a number or a quote.
- Anything sent in someone's name says whose name it is.
- It will refuse to build a satisfaction filter, a star gate, a smiley face router or any survey whose answer changes where the customer is sent. That is review gating, prohibited by name in Tripadvisor's guidelines and by description in Google's, and it is the practice the CMA points at when it describes encouraging only the satisfied to leave reviews. There is no compliant version to design instead, so the skill says no and stops.
- Never write, edit, translate, tidy or suggest the words of a review, and never stand over somebody while they write one.
- Never ask a member of staff, a family member, a friend, a supplier or a fellow trader to write one, however genuine their visit was. The platforms exclude them regardless of honesty, and a removed review takes the listing's credibility with it.
- Never state a response rate, a review conversion rate, an industry average or a target number of reviews per month. No published figure exists for a single independent UK venue, the numbers in circulation come from companies selling review software, and an owner who sets a target from a borrowed number starts filtering to hit it.
- Never send a review request to somebody whose complaint is not yet resolved. It reads as a request to be quiet, and the CMA warns against making dispute resolution contingent on a consumer not leaving a negative review.
- This output is a working document prepared for the owner to check against their own customer records, their booking system's terms and the current published policies of the platform they use, and for their solicitor or trade body to check where an incentive, a contract term or a data protection question is involved. It applies published rules and platform policies to a draft message; it is not legal advice and it is not confirmation that a send complies.

## Built from
- Digital Markets, Competition and Consumers Act 2024, Schedule 20, paragraph 13, https://www.legislation.gov.uk/ukpga/2024/13/schedule/20/paragraph/13/enacted, Act of Parliament given Royal Assent in 2024, read 16 September 2026: the banned practice, the definitions of a fake consumer review and of a concealed incentivised review, and the misleading publication examples, behind steps 1, 3 and 5.
- The Digital Markets, Competition and Consumers Act 2024 (Commencement No. 2) Regulations 2025, SI 2025/272, https://www.legislation.gov.uk/uksi/2025/272/made, made 2025, read 16 September 2026: confirmation that Chapter 1 of Part 4 came into force on 6 April 2025, which is why step 1 treats this as live law.
- Competition and Markets Authority, "Fake reviews" (CMA208), https://assets.publishing.service.gov.uk/media/67eeb64fe9c76fa33048c790/CMA208_-_Fake_reviews_guidance.pdf, published 4 April 2025, read 16 September 2026: the cherry picking passage at 4.5 in step 1, the interference list at 4.4 in step 7, and the incentive conditions at 3.7 in step 3.
- Google, "Prohibited and restricted content", Maps User Contributed Content Policy, https://support.google.com/contributionpolicy/answer/7400114, no publication date shown on the page, read 16 September 2026: the selective solicitation ban in step 2, the staff target and on premises pressure rules in step 4, and the permitted form of asking in step 8.
- Tripadvisor, "Trust and safety review posting guidelines", https://www.tripadvisor.co.uk/Trust-lvBd3L1aU38Y.html, no publication date shown on the page, read 16 September 2026: the review gating prohibition in step 2, the incentive and employee incentive bans in steps 3 and 4, the bias list in step 5, and the one year window in step 7.
- Information Commissioner's Office, "Identify direct marketing", https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/direct-marketing-guidance/identify-direct-marketing/, latest update shown on the page 20 August 2025, read 16 September 2026: the definition and the service message test in step 6.
- The Privacy and Electronic Communications (EC Directive) Regulations 2003, regulation 22, https://www.legislation.gov.uk/uksi/2003/2426/regulation/22/made, SI 2003/2426, read 16 September 2026: the three conditions of the soft opt in quoted in step 6.
- Competition and Markets Authority, "Fake and misleading reviews: 5 businesses under CMA investigation", https://www.gov.uk/government/news/fake-and-misleading-reviews-5-businesses-under-cma-investigation, published 27 March 2026, read 16 September 2026: the Dignity line and the 10% of global turnover penalty in step 5.
- Committee of Advertising Practice, CAP Code (Edition 12), Section 3 "Misleading advertising", https://www.asa.org.uk/type/non_broadcast/code_section/03.html, no publication date shown on the page, read 16 September 2026: rules 3.47 and 3.50 on evidence, contact details and permission, which set what step 9 logs.

# review-ask

## You are given
A customer list exported from a booking system, till, PMS or spreadsheet, covering one stated period, with one row per transaction. The written selection rule that produced it: the start date, the end date, the transaction type, and the count the owner expects. The platform the reviews are being asked for, one of `google`, `tripadvisor`, `facebook`, `own site` or `other`, with the name of the platform's published policy page and the date it was read. The final wording of every message variant, exactly as it will be sent, each with its channel (`email`, `sms`, `whatsapp`, `printed card`, `qr code`) and its sender name. The incentive decision already made by Claude: whether an incentive is offered, what it is, the disclosure wording, and the sentence of platform policy that permits or forbids it. The lawful basis or PECR route already chosen for each channel, and the opt out wording. A staff and affiliate list: employees, former employees, family members, named suppliers, and any trading address of a business the owner also owns. Any previous send log.

## Produce
Write into a `./review-ask-output/` folder:

1. `recipients.csv` with these columns in this order: `row_ref`, `transaction_date`, `transaction_type`, `contact_channel`, `contact_held`, `in_selection_rule`, `excluded_reason`, `previously_asked_date`, `complaint_open`, `affiliation_flag`. `row_ref` is R0001 upward. `in_selection_rule` is `yes` or `no`. `excluded_reason` is exactly one of `outside date range`, `wrong transaction type`, `no contact details held`, `opted out`, `asked already for this visit`, `complaint open`, `affiliated person`, `not excluded`. `complaint_open` and `affiliation_flag` are `yes` or `no`.
2. `message-set.csv` with columns: `variant_ref`, `channel`, `sender_name`, `message_text_verbatim`, `characters`, `link_url`, `link_destination_count`, `contains_star_word`, `star_words_found`, `contains_content_request`, `content_requests_found`, `contains_offer`, `opt_out_present`, `opt_out_text_verbatim`. `contains_star_word` searches for the literal tokens `star`, `stars`, `5`, `five`, `positive`, `great`, `glowing`, `lovely`, `honest`, `good` and records every hit with its position. `link_destination_count` counts distinct URLs across the whole set and must be 1.
3. `gating-check.csv` with columns: `variant_ref`, `question_asked`, `answer_options`, `destination_per_answer`, `destinations_identical`, `verdict`. One row per branching point found in the supplied messages, surveys, landing pages or QR flows. `destinations_identical` is `yes` or `no`. `verdict` is `pass` or `review gating found`. Any `review gating found` row is also listed in `gaps.md` as the first item.
4. `incentive-check.csv` with columns: `incentive_offered`, `incentive_description`, `depends_on_rating`, `disclosure_wording_verbatim`, `platform`, `platform_policy_url`, `platform_policy_quote_verbatim`, `platform_permits`, `law_permits_with_disclosure`, `outcome`. `platform_permits` and `depends_on_rating` are `yes` or `no`. `outcome` is exactly one of `no incentive offered`, `permitted and disclosed`, `removed because the platform prohibits it`, `removed because it depends on the rating`.
5. `affiliation-screen.csv` with columns: `row_ref`, `name_as_supplied`, `matched_against`, `match_type`, `action`. `matched_against` is exactly one of `employee list`, `former employee list`, `family or friend list`, `supplier list`, `owner's other business`, `same category within 16 kilometres`, `no match`. `action` is `removed from list` or `kept`.
6. `send-log.csv` with columns: `send_date`, `selection_rule_verbatim`, `period_start`, `period_end`, `recipients_count`, `platform`, `variant_ref`, `channel`, `sender_name`, `incentive_outcome`, `lawful_basis_or_pecr_route`, `opt_out_method`. Appended to any previous log, never overwritten.
7. `gaps.md` - a numbered list of: any review gating verdict; any message containing a star word, a rating, an offer or a content request; any second link destination; any recipient with no recorded lawful basis or PECR route; any message with no opt out; any recipient with an open complaint or an affiliation flag that was not removed; any row whose `in_selection_rule` is `yes` but which the supplied count does not account for; and any platform policy quoted with no URL or no read date.

## Rules
- Codex counts, matches, checks and records. It never writes a message, never rewords a supplied one, never invents a selection rule, and never makes the incentive or lawful basis judgement that was supplied to it.
- Never write, draft, translate or suggest the text of a review, a rating, or anything a customer might paste into a review box.
- Never build, describe or document a flow in which the answer to a satisfaction question changes the destination. Where one is found in the input, record it in `gating-check.csv`, list it first in `gaps.md`, and write nothing that implements it.
- Never mark `platform_permits` as `yes` without a verbatim quote from the named policy page and the date that page was read, both recorded in the row.
- Never write a response rate, a review conversion rate, an industry average, a benchmark or a target number of reviews into any file. No such figure exists for a single independent UK venue.
- Counts are counted, never estimated. Character counts include spaces and punctuation. Token searches are literal, case insensitive, and record the position of every hit rather than a yes or no alone.
- Every quoted message, policy line and opt out is copied exactly, including punctuation, capitalisation and any emoji.
- Never send anything, never connect to an email, SMS or messaging platform, and never call a review platform's API.
- A recipient removed for any reason stays in `recipients.csv` with the reason recorded. Rows are never deleted, because the removed rows are the evidence that the selection rule was applied.
- Use British English, GBP and DD Month YYYY dates. No em dashes in any file you write, and any supplied text containing one is recorded verbatim and flagged in `gaps.md`.
- Every file ends with this line: this is a working document prepared for the owner to check against their own customer records and the current published policies of the platform they use, and for their solicitor or trade body to check any incentive, contract or data protection question. It applies supplied rules to a draft send and is not confirmation that the send complies.

## Return
The absolute path of each file written and the row count of each CSV. The selection rule verbatim with its period start and end. The number of rows in the supplied list, the number where `in_selection_rule` is `yes`, and the count of each `excluded_reason`. The number of message variants, the distinct link destination count, and every star word, content request and offer found, each with its variant reference and position. The `gating-check.csv` verdict for every branching point, stated in full. The `incentive-check.csv` outcome with the platform policy quote that decided it. The number of rows removed by the affiliation screen and which list each matched. The number of recipients with no lawful basis or PECR route recorded, and the number of variants with no opt out. The new `send-log.csv` row exactly as written, and the `gaps.md` item count.