4.6 stars. Great. What do I do with that on a Tuesday?
We pulled 2,561 review texts for one brand and grouped them by what people were actually complaining about, and the top theme was not the product.
It was the case. The little box it comes in.
Their most repeated complaint was about the thing they probably spent the least time on.
That is a product roadmap item for them and a positioning line for you, and it was sitting in public the whole time, spread across two and a half thousand reviews that nobody was going to read end to end.
The reason this works is that complaints repeat and praise does not. Happy people write "love it, great product" in nine different ways. Unhappy people are specific, because they are annoyed, and annoyance is detailed.
Two things we are strict about, both learned the hard way.
Store reviews and Trustpilot reviews never get added together. A review on the brand's own product page is a review of a product, and the brand moderates it. A Trustpilot review is a review of the company: delivery, support, what happened when someone tried to cancel. Those are two different questions, and averaging them produces a number that answers neither.
You see why it matters when the two disagree. In one head to head, the competitor won the AI answers comfortably, 35% against 16%. Then we looked at Trustpilot: 2.07 stars from 15 reviewers, against 4.26 from 129. The engines were recommending the brand that people were angriest at.
Second thing: for B2B, the raw material is the cons field on G2. Not the review body, the cons field specifically, because reviewers use it as a list. Read a few hundred of them and you have an objection library written by your competitor's own customers, in their words, which is the only wording that survives contact with a sales call.
What we cannot do is verify a reviewer. We read what is published. Fake reviews exist and we do not detect them, and a theme built on twelve reviews is a hint, not a finding.
One thing to do next: open your competitor's one star reviews. Not the two, not the three. The ones.
Read twenty of them and write down every noun that appears twice.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.