Review Moderation

Also known as review approval, review queue

Every review app ships a queue and a hide button, and the button does not know the difference between a competitor's spam and a legitimate one star. Nothing in the interface distinguishes them, which means the policy has to, and the policy has to exist in writing before the queue contains something you would rather nobody read. A moderation process is not an admission of anything. A moderation process invented on the afternoon a bad review arrives is a gate with extra steps.

The defensible list is short and worth writing down in advance. A review is removable when it did not come from a real customer, when it is a competitor placement or contains a link to another store, when it is a duplicate or a test submission, when it is obscene, harassing, or discriminatory, when it exposes personal data, and when it is a review of a different product. That last one has an edge that stores get wrong in their own favour: a complaint about slow delivery is about your store even though it is not about the product, and burying it as off topic is the same move as gating. Genuinely off topic means a review of something the customer bought elsewhere. The item that is not on the list, and never joins it, is that the review is negative.

The FTC has stated that selectively suppressing negative reviews is deceptive, and the rules aimed at review suppression turn on misrepresentation: the offence is presenting a review display as what customers said while the unflattering ones have quietly been taken out. Read that way, having a queue is not the problem and removing a review full of somebody's phone number is not the problem. Removing a review because it costs you sales is exactly the problem, whatever reason is entered in the log.

Consider a Shopify store selling scented candles, with twelve reviews sitting in the queue on a Monday morning. Six unremarkable four and five stars go up untouched. The other six need a decision. A one star saying the jar arrived cracked with the outer box intact is published, and then answered, because it is a real customer describing a real fault. A five star that includes the buyer's full name, street address, and phone number, pasted in from their order confirmation, is held and the customer asked to resubmit without it; the rating and the text stay, the personal data does not. A review about a diffuser the customer bought from another store is removed as off topic, with the reason logged. A one star that is four sentences of abuse aimed at a named packer is removed, and here the hard part is that it also contains a genuine complaint about a leaking lid, so the right move is to strip the abuse or ask for a rewrite rather than to lose the complaint with it. A five star from an email that has never ordered, carrying a discount code for another shop, is removed as spam. And the sixth, a two star saying the scent throw is weak in a large room, is published, because it is accurate. That one is answered, and the room-size guidance goes into the product description, which is where it should have been in the first place.

Process matters more than judgement here, because judgement drifts under pressure. Publish the policy where a shopper can read it. Read the text before the star, so the decision is made on content and not on the number. Log every removal with a reason, and keep the removed reviews rather than deleting them, so you can show your work if anyone asks. Never edit review text silently; fixing a typo and removing the sentence about the broken zip are the same action from the outside. And whatever average you display has to be computed on what you actually published under the policy you can state out loud.

The timing choice is between holding everything for approval and publishing on arrival with a check afterwards. Pre-moderation catches spam before a shopper sees it and adds a delay that makes your review dates less useful, plus a daily temptation. Post-moderation removes the temptation and exposes you to spam for a few hours. The workable middle is to auto-publish, filter automatically for links and known spam patterns, and put a human on whatever the filter flags. A queue that is only ever emptied when a review is unflattering is not moderation, whatever the process document says.

One asymmetry to plan around: on your own storefront you control the display, but on Google, Trustpilot, or a marketplace you can only flag a review for a policy breach and the platform decides. Any strategy that assumes you can clean up after a bad week is only true on the surface you own, which is the smaller half of your reputation.

There is an AI search consequence, and we should be plain about our own interest in it, because we sell review software. Answer engines corroborate across sources. A storefront that is uniformly positive while third-party sources carry the same recurring complaint reads as curated rather than good, and the complaint surfaces in the summary anyway. Moderating to a stated content policy leaves you with a mixed corpus that matches what the rest of the web already says, which is the condition under which corroboration works in your favour. Take that as an argument if it holds on its own; the test that does not depend on who is making it is whether you could publish your moderation policy and your removal log side by side without wanting to explain anything.