review_generation v1.0.0

name: review_generation description: "Get real customer reviews ethically: timing the ask after a good outcome, picking the right channel per platform, ask templates for email, SMS, in person, and in product, and response playbooks for good and bad reviews. Use when someone wants more Google or product reviews, asks about review gating or incentives, needs review request copy, or needs help replying to a negative review without breaking FTC or platform rules."

Review Generation

Purpose

Build a steady flow of honest reviews from real customers. The work has four parts: ask at the right moment, ask on a channel the platform allows, make the ask take under a minute, and reply to every review that lands. Shortcuts that distort the rating (gating, incentives tied to sentiment, fake or insider reviews without disclosure) are banned by the FTC Consumer Reviews and Testimonials Rule and by platform policy, and they get reviews removed.

Workflow

1. Check the platform rules first. The same ask is legal on one platform and a violation on another. Google allows asking every customer, with no incentives and no screening. Yelp does not want businesses to ask at all, and its software moves solicited reviews out of the rating. Split the flow by platform before writing any copy. Full matrix in references/platform-rules.md.
2. Pick the trigger, not a date on a calendar. Ask after a verified good outcome: job completed, delivery received and used, milestone reached, support case resolved, repeat purchase made. Never ask at checkout, at signup, before delivery, or while a problem is still open. Timing ranges by business type in references/timing-and-channels.md.
3. Ask everyone the same way. Send the identical neutral ask to every customer who hits the trigger. Do not pre screen with a private satisfaction question and then send only happy customers to the public form. That is review gating. A private feedback route can exist, but the public review link goes to everyone, in the same message, with the same prominence.
4. Write the ask short and neutral. One sentence on the experience, one direct link, no requested rating, no suggested wording, no reward. Templates per channel in references/ask-templates.md. Ask once, send at most one reminder, then stop.
5. Make it easy. Link straight to the review form, not to a homepage or profile page. Use a QR code on receipts, packaging, and invoices for in person businesses. Test the link logged out before any campaign ships.
6. Reply to every review. Thank positive reviewers by name and mention the specific thing they praised. For negative reviews: acknowledge, state facts briefly, apologize for the experience without admitting things that did not happen, offer a real phone number or email to continue offline, and stop. Never offer anything in exchange for changing or removing a review. Playbooks in references/response-playbooks.md.
7. Handle policy breaking reviews through the platform. If a review is fake, off topic, from a competitor or employee, or contains private data or hate, flag it in the platform tool with the policy it breaks. Do not contact the reviewer with pressure, threats, or legal claims.
8. Measure the flow, not just the rating. Track asks sent, reviews received, response rate, time to reply, and rating distribution by month. A sudden spike in five star reviews with similar wording is a warning sign, not a win. Use bin/review_log_check.py to sanity check a CSV export of asks and reviews.

Output Contract

When this skill produces a review program, request copy, or a reply, the deliverable includes:

• Trigger and timing plan: the outcome that triggers the ask, the delay after it, the channel, and the reminder rule, stated per product or service line.
• Platform split: which platforms get an active ask, which get passive treatment only (Yelp), and the rule that drives each choice.
• Ask copy: final wording per channel, checked against the banned list in the Operating Rules, with the direct review link placed.
• Response drafts: for any live review, a reply that follows the playbook for its rating band, ready to post, with no offer tied to editing the review.
• Compliance note: any incentive, insider review, or testimonial reuse in the plan, with its disclosure or the reason it was cut.

Operating Rules

1. No fake reviews, ever. No writing, buying, selling, or posting reviews from people who did not have the experience, including AI generated ones. Banned under 16 CFR Part 465, effective October 21, 2024.
2. No incentive tied to sentiment. Never offer money, discounts, gifts, entries, or anything of value for a positive review, a five star rating, or for changing or removing a negative review. Google bans incentives for any review at all, so the safe default is no incentives, period.
3. No review gating. Do not screen by happiness, stars, or a survey score and route only happy customers to the public form. Ask everyone who hits the trigger, the same way.
4. No suppression. Do not hide negative reviews on your own site, bury them with display tricks, or use threats and unfounded legal claims to get one taken down. Form contracts cannot ban or penalize honest reviews either: the Consumer Review Fairness Act voids those clauses.
5. Insiders disclose or do not post. Employees, owners, family, and paid testers do not review the business on Google. Anywhere an insider testimonial is lawfully used, the connection is disclosed clearly, next to the testimonial.
6. Never script the customer. No suggested star rating, no draft text for them to paste, no list of points to mention, no request to name a staff member.
7. One ask, one reminder, then stop. Repeated asks, on premises pressure, and staff quotas for collecting reviews break platform policy and produce pushy reviews that mention the pressure.
8. Reuse needs permission. Quoting a review in ads or on a product page needs the reviewer permission where the platform terms require it, and the quote must match what they actually wrote. A review is not a testimonial until that step is done.
9. Timing ranges in the references are starting points, not guarantees. Test the delay against real response rates for that business and adjust. Never present a timing range or a response rate as a proven fact for a business that has not measured it.