Reviews, Forums and Videos: What Customers Say Elsewhere

A product page presents what the brand wants to highlight. Customers say other things, in reviews, on forums and in the comments under videos. They describe how they really use the product, what surprised them and what almost stopped them from buying. That is product knowledge. It is also the easiest kind to misuse.

What a product page leaves out

Product copy is written before most customers have used the product. It describes features, materials and intended uses. It rarely answers the questions that come up later. Skincare shows the pattern clearly, see the skincare questions a product page rarely answers.

Can a 16-inch laptop fit in this bag with its sleeve? Does the paint cover a dark wall in one coat? How does the strap feel after a few months? These questions come back again and again in public discussions, often long after the product page was written.

For a merchant, these conversations show three things. The questions customers ask before they buy. The uses the brand did not plan for. The objections that cost sales or cause returns. An AI advisor that ignores them will answer the questions the brand expected, not the ones customers actually ask.

Keep every piece with its source

A sentence taken from a forum, with no record of where it came from, cannot be checked, corrected or withdrawn. Every piece of evidence should keep its source, its date, its language and the product it refers to.

Context matters as much as content. One review is one person's experience. A pattern needs several independent accounts. An old review may describe a version of the product that has since changed. Without the source, none of this can be judged.

Keep customer voice apart from brand voice

Brand copy is real knowledge. It is simply not a customer speaking. When the two are mixed, a customer can no longer tell an experience from a promise. A marketing claim starts to sound like a customer's verdict, and one customer's complaint starts to sound like a fact.

EU consumer law already draws this line for reviews. Since 28 May 2022, the Unfair Commercial Practices Directive, as amended by Directive (EU) 2019/2161, requires a trader that gives access to consumer reviews to say whether and how it checks that they come from people who actually used or bought the product. Fake reviews are banned. So is misrepresenting reviews, for example by showing only the positive ones.

The same discipline makes sense for an AI advisor. When it says “customers often mention”, that statement should rest on real accounts, kept with their source, and the selection should not make a part look like the whole.

The merchant approves before the advisor speaks

Collected evidence is a proposal, not an answer. The merchant reviews it, corrects it or rejects it. Only approved knowledge should reach a customer, and it should keep its source once approved.

Understanding a product from public discussions is also different from republishing someone's words. Before quoting a review or a video, check the platform's terms and the rights of the person who wrote or filmed it.

Criticism is worth keeping

The tempting move is to delete what hurts. “The zip is hard to open with one hand.” “The color looks darker than in the photos.” Deleting these comments throws away the most actionable part of the evidence.

Not every criticism belongs in front of a customer who is being helped to choose. But it belongs in front of the brand. A recurring complaint can reveal a missing fact, such as sizing or compatibility guidance. It can reveal a product defect, or a page that sets the wrong expectation. Sometimes a frank limitation also helps the right customer choose correctly and avoids a return.

The principle is simple: classify, never discard. Each critical comment gets a label and goes to the team that can act on it. The merchant then decides whether an advisor should mention it, and how.

How Querytail approaches this

Sourcing Agent collects product information from a merchant's catalog, technical documents, customer reviews and customer conversations, as well as public discussions on the web, in forums and in videos, and records where each piece comes from. Product Engine turns it into suggested enrichments, which the merchant's team approves, edits or rejects in Merchant Console before any use.

Each piece of evidence also records who is speaking. Customer experience and brand copy stay distinct, and criticism is routed to the brand rather than discarded.

FAQ

Can an AI advisor rely on customer comments? Yes, when the source is kept and the merchant has approved the use. Quoting a person's words is a separate question, so check the platform's terms and the author's rights first.

Should we remove negative reviews from product knowledge? No. Keep them, labelled, and route them to the team that can act. The merchant decides what reaches customers.

Is one review enough to state a product fact? No. One review is one experience. A product fact needs a reliable source, and a pattern needs several independent accounts.

Want to see how this applies to your catalog? Request a demo.