How to automate product review requests: the rules on asking

Asking customers for reviews is straightforward to automate and is governed by more rules than most people realise. The major marketplaces and review platforms each prohibit offering anything in exchange, and most prohibit asking only your happy customers — and enforcement ranges from removing the reviews you gathered to suspending the listing.

Within those rules there is still a lot of room, and the thing that determines how much you get is timing. Asking at the wrong moment is the difference between a response rate that justifies the automation and one that does not.

What we'll cover

• What the platforms prohibit
• Gating, which is the common mistake
• Timing is most of the result
• Building it with CodeWords
• Who to ask, and who not to
• The request itself
• Which platform to send them to
• Responding to what comes back
• Making it survive
• Limits worth knowing about
• What to build first
• Frequently asked questions

What the platforms prohibit

Incentives of any kind — discounts, entries into a draw, loyalty points, free products — in exchange for a review. This is near-universal across marketplaces and review platforms.

Asking only for positive reviews, explicitly or by implication in the wording.

Filtering who gets asked based on their likely sentiment, which is the gating problem below.

Writing or editing the review on a customer's behalf, in any form.

Reviews from staff, family or anyone with an interest, which several platforms treat as a serious violation.

Rules differ in detail between platforms and some allow a neutral incentive for leaving feedback of any kind, so check the specific terms rather than assuming a general rule.

Gating, which is the common mistake

The pattern is familiar: ask how the experience was, send happy customers to a public review site and unhappy ones to a private form.

It is prohibited by the major review platforms and marketplaces, and it has been the subject of regulatory action in several markets.

It is detectable from an implausible rating distribution, which is one of the things platforms look for.

The penalties are real — review removal, listing suspension, and in some cases public regulatory findings.

It also damages the reviews you keep. A five-star average that everybody knows is filtered persuades nobody.

What is allowed is asking everybody and handling the resulting negative reviews well, publicly.

Routing a complaint to support is fine — provided the review request went to everybody regardless.

Timing is most of the result

Ask after the value has been experienced, not after the transaction completed, which are usually different moments.

For physical goods that means after delivery plus enough time to have used the item, which varies enormously by product.

For services, after the outcome, which may be weeks later.

For software, after a success event rather than after a period, since a person who has not yet got it working has nothing to say.

Too early gets a review of the purchase experience, which is not what the listing needs.

Too late gets nothing, since the experience has faded. Test the window against your own response rates rather than adopting a default.

Building it with CodeWords

CodeWords connects to more than 3,000 integrations, so the order or product system, the delivery data and the messaging channel can meet in one automation.

1. Open CodeWords and start a new automation.
2. Describe what should happen in plain language to Cody, the automation builder: the trigger event, the delay, the wording, and who is excluded.
3. Authorize the connections — the store or product system, the delivery or fulfilment source, and the email or messaging platform.
4. Describe the exceptions: an open support ticket, a return, a refund, somebody asked recently, a gift order.
5. Run it against last month's orders as a dry run and read the list of who would have been asked.

You describe the outcome; Cody builds it, connects it, and deploys it. The free plan covers light use, with Pro at $39 per month and Business at $100 per month as usage grows; details are on the pricing page.

Who to ask, and who not to

Everybody who received the product or service, which is the rule that keeps this legitimate.

Not anyone with an open support ticket, which is a courtesy rather than a filter — resolve it first and ask afterwards.

Not returns and refunds, where the request is straightforwardly inappropriate.

Not gift recipients or purchasers where you cannot tell who actually used it.

Not more than once, per purchase, with a global cap on how often any customer is asked across products.

Not during an incident affecting that customer, which is a judgement worth encoding rather than leaving to chance.

The request itself

Short. This is a favour being asked and the length should reflect that.

Neutral wording. "How did you find it" rather than anything that presumes an answer, which is both a rule and better practice.

One link, direct to the review form, with as few steps as possible after the click.

Say how long it will take, honestly, since the perceived effort is what stops most people.

Explain why it helps, briefly and specifically — other customers deciding, not your business metrics.

From an address that accepts replies, since a proportion of people will reply with feedback rather than posting it, and that is worth having.

Which platform to send them to

Where your buyers actually look, which differs by category and is worth establishing rather than assuming.

Marketplace reviews where you sell on a marketplace, since that is where the purchase decision happens.

Do not split the ask across several platforms in one message, which reduces the response on all of them.

Rotate deliberately if you need coverage on several, rather than asking everybody for everything.

On-site reviews are yours and are useful, and they carry less weight with buyers than independent ones.

Check each platform's rules separately, since what is permitted on one is prohibited on another and the automation needs to know which it is sending to.

Responding to what comes back

The request is half the system. What happens to the reviews afterwards is what prospective buyers actually read, and it is usually handled by nobody in particular.

Respond to the negative ones publicly, specifically, and without defensiveness. A well-handled complaint persuades more than an unbroken run of praise.

Respond quickly. A month-old unanswered complaint reads as a company that is not paying attention, whatever the answer would have been.

Do not respond identically. Templated replies to every review are visible as templates and undermine the individual ones.

Route them by content, not by rating, since a four-star review describing a specific defect is more actionable than a one-star review with no detail.

Feed them to the people who can fix things, which for most companies is the step that never happens — reviews are read by marketing and describe product problems.

Notice the pattern. Three reviews mentioning the same thing is a product signal, and it is a faster one than most internal reporting.

Making it survive

Make it idempotent so nobody is asked twice for one purchase.

Check the current state before sending — a refund or ticket opened after scheduling should stop it.

Respect a global contact frequency cap across all your automated messaging.

Handle delivery data gaps, since a missing delivery confirmation should delay rather than trigger the request.

Report the outcome: requests sent, suppressed and why, and the response rate by timing window.

Limits worth knowing about

Response rates are low, in single digits for most categories, which is normal rather than a sign of a broken automation.

Platform APIs for review status vary, and many do not tell you who reviewed, so attribution is partial.

Consent rules apply to review requests as commercial messages in some jurisdictions.

Marketplaces often have their own request mechanism and may prohibit contacting buyers outside it.

You cannot remove a negative review you dislike, and attempting to have legitimate reviews removed is itself a risk.

What to build first

A single request, triggered by the moment the customer has actually experienced the product rather than by the order, sent to everybody. No incentive, no filtering, neutral wording. It is the compliant version, it takes an afternoon, and the timing decision is the one that determines the response rate more than anything else you could change.

Two habits make the difference. Ask everybody rather than pre-selecting, since gating is prohibited, detectable and damaging to the reviews you keep. And measure the response rate by timing window, because the right delay differs enormously by product and the default in whatever tool you are using was not chosen for yours.

Frequently asked questions

Can I offer a discount in exchange for a review?

No. Incentives of any kind are prohibited across the major marketplaces and review platforms, and the penalties include removing the reviews you gathered and suspending the listing.

What is review gating and why is it a problem?

Asking how the experience was and routing happy customers to a public review site while unhappy ones go to a private form. It is prohibited, detectable from the rating distribution, and has been the subject of regulatory action in several markets.

Can I send complaints to support instead of a review site?

Yes, provided the review request itself went to everybody regardless of sentiment. Handling a complaint well is fine; deciding who gets asked based on expected sentiment is not.

When should I ask for a review?

After the value has been experienced, not after the transaction. For physical goods that means after delivery plus time to use the item; for services after the outcome; for software after a success event rather than a fixed period.

Who should be excluded?

Anyone with an open support ticket, returns and refunds, gift orders where you cannot tell who used the product, anyone asked recently, and anyone affected by an ongoing incident.

What should the request say?

Something short and neutral — "how did you find it" rather than anything presuming an answer — with one direct link, an honest estimate of how long it takes, and a specific reason it helps other customers.

What response rate should I expect?

Single digits for most categories. That is normal rather than a sign the automation is broken, and the timing window matters more than the wording.

Can I get a negative review removed?

Not because you dislike it, and attempting to have legitimate reviews removed is itself a risk. Responding well publicly does more for prospective buyers than a missing review would.

Should I respond to negative reviews?

Yes, publicly, specifically and quickly. A well-handled complaint persuades prospective buyers more than an unbroken run of praise, and an unanswered month-old complaint reads as a company not paying attention.

Who should read the reviews internally?

The people who can fix what they describe. Reviews are usually read by marketing and usually describe product problems, and three mentioning the same thing is a faster signal than most internal reporting.

Should I ask for reviews on several platforms at once?

Not in one message — splitting the ask reduces the response on all of them. Rotate deliberately across platforms if you need coverage on more than one, and check each platform's rules separately since they differ.

Related reading

• Automating social proof collection
• Automating sentiment analysis on reviews
• Automating abandoned cart emails