TikTok comments aren’t reviews. Read them anyway.

August 23, 2026 · 7 min read

Read them, but not as reviews. A comment section has no purchase verification, no rating and no way to separate a buyer from a bystander, so the amount of praise under a video is not a score. Read it instead for the one thing a review section rarely produces: a specific complaint, from someone with nothing to gain, arriving long after everyone stopped paying attention.

Here is what to weight, what to throw out, and why your own fake-comment detector is worse than you think it is.

A comment is not a review, structurally

TikTok Shop reviews live on the product page and come through a narrow door. The order has to be marked delivered before the buyer can write one, and they get 60 days from that point. Whatever else you think of them, the person typing bought the thing and received it.

The comment section under the video has no door. Anyone types. Somebody who owns it, somebody who wants it, somebody who is only here for the sound, and somebody who was paid a few cents. Same weight, same font, no badge.

That does not make comments worthless. It makes them a different instrument. Reviews are a filtered, verified, narrow record. Comments are an unfiltered, unverified, enormous one. You reach for them for different jobs.

Top comments are ranked for engagement

Sorted by top, a comment section is a popularity contest. The winner is a joke. Second is “link?”. Third is somebody telling the creator she looks like an actress. None of it is about the product.

The comment you actually needed — “bought it, week three, the motor started clicking” — landed on day forty, got eleven likes, and sits four hundred comments down.

So switch the sort to newest. We have said that before for a different reason: the creator's late answer about the link lives down there too. The reason that matters here is simpler. On day one, nobody in the comments owns the product. On day thirty, some of them do.

What comments do better than reviews

One category, mostly: the failure that happens outside the review window, or in a place the review form never asks about.

• The delayed failure. A review written inside 60 days of delivery cannot describe month four. A comment can, and often does, because the person went back to the video that sold it to them.
• The thing nobody puts in a review. Smell. Noise. Whether a fabric photographs like it looks in a room. Whether a self-tanner stains a pillowcase.
• Fit told against a body.“I'm 5'2 and it hits mid-calf” is worth more than a star rating, and it appears in comments far more often than in reviews.
• The correction. Somebody naming the identical unbranded version, or pointing out the brand relaunched the same item under a new name. A review on a product page has no reason to send you somewhere else. Commenters do it constantly.

Notice the pattern. Every one of those is specific, and most are inconvenient to somebody. That is not a coincidence. It is the whole heuristic.

The “broke me out” pattern

One person saying a moisturizer broke them out is not information. Skin is individual, and that person may have changed four things that week.

The same complaint from six unrelated accounts, in six different phrasings, with six different details, is information. Not proof. Information.

The test is not how many people said it. It is whether they said it differently. Coordination produces agreement in wording. Reality produces agreement in substance and variation in wording. Ten comments reading “this broke me out” word for word is one voice ten times. Ten comments reading “purged for two weeks”, “tiny bumps along my jaw”, “fine on my cheeks, not my chin” is ten people.

How comments get seeded

Bought comments are a retail product. Vendors sell TikTok comment packages openly, from a couple of dollars for ten of them up to orders in the thousands, and the tier that matters here lets the buyer write the exact text. If a comment section has ever read to you like it was drafted by the brand, sometimes it was.

Both the platform and the law are against it. TikTok's community guidelines prohibit fake engagement, bought comments included, and the platform strips the inflated metrics and restricts the accounts behind them. In the US, the Federal Trade Commission's Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024 and made buying and selling fake reviews and testimonials, along with fake indicators of social media influence, grounds for civil penalties. TikTok's own reviews policy requires an incentivised review to be disclosed as one.

The limit, said out loud: none of that means the comment section is clean. It means the crude version got riskier to buy. What survives enforcement is the careful version, and the careful version is precisely the one you cannot spot.

You are worse at this than you think

Most advice on this topic stops at a list of tells. It should not, because the tells do not work as well as they read. When researchers put genuine and fabricated reviews in front of people and ask them to sort the two, accuracy across published studies lands roughly between 50 and 65 percent. That is a coin flip, to slightly better than a coin flip, and coaching people on what to look for moves it very little.

So stop running a detector. Run a filter instead. You are not trying to identify who is lying. You are trying to change what you count.

The tells that survive being wrong half the time:

• Repeated wording. Identical or near-identical sentences under different names. The strongest single signal, because bought custom comments are one text pasted many times.
• Burst timing. A cluster of glowing comments all posted within the same short window, especially weeks after the video went up and nowhere near its traffic peak.
• Praise with no object.“Obsessed.” “Need this.” “Adding to cart.” Somebody who owns a thing mentions the thing.
• Excitement about the deal, not the item. Comments about how cheap it is, from people who never say what it does.
• An account with no history. This one is weak, so weight it accordingly. Enormous numbers of real people watch for years and comment once.

Two minutes, in order

• Sort by newest. Thirty seconds, and it moves you from the audience to the buyers.
• Look for negatives first. They are rarer, they cost the writer something to post, and nobody pays for them.
• Keep only the specific ones. A detail an owner would know. Discard every generic verdict in both directions, including the enthusiastic ones.
• Check whether the agreement varies. Same substance, different words, unrelated accounts. That is consensus. Same words is one voice.
• Take it to the product page. Reviews there are gated on delivery, so use them for the opposite question: did it arrive, and was it the item in the video?
• When the two disagree, believe the specific one. Not the positive one, not the negative one. The one carrying a detail.

Where this leaves you

Two minutes of reading a comment section properly beats twenty minutes of reading it the way it is sorted by default. It still will not tell you whether a serum works, and it never will, because the comment section is a sample of people who were excited enough to type.

What it will tell you is what goes wrong, and roughly how often, from people who had no reason to mention it.

The harder part usually comes before any of this. You cannot read the comments on a product until you know which product it is, and the video did not say. That part is what Stealsy does. Share the video to the app and it identifies the products, prices them, links each one to a shop page it has actually opened and checked and shows the cheaper dupes it can find. It does not score comment sections or grade sentiment for you. It gets you to the right product page fast enough that you still have the patience to read what people said about it.

Get Stealsy for iPhone — share a video, get the products, prices and verified links.