Review platforms are third-party sites where customers rate products and companies — G2, Capterra, Trustpilot, or Yelp. According to an SE Ranking study, they accounted for 3 of the 5 most-cited domains in AI Overviews, even though their organic traffic dropped 76.5 to 92.2 percent since the start of 2024.
Those two numbers side by side are the whole story of this article. Sites that fewer and fewer people visit carry more and more weight in AI answers. First why, then — more importantly — what to do about it.
What the study actually measured
SE Ranking analyzed 22,729 AI Overviews across 30,000 commercial keywords in the US (a December 2025 snapshot, 23 review platforms). Two headline findings:
- Review platforms made up just 8.5 percent of all links in AI Overviews — but 3 of the 5 most-cited domains were review sites.
- Over the same period, those platforms lost most of their organic traffic between early 2024 and late 2025.
| Aspekt | Share of review-site citations | Change in organic traffic 2024–2025 |
|---|---|---|
| Gartner Peer Insights | 26.0% | −76.5% |
| G2 | 23.1% | −84.5% |
| Capterra | 17.8% | −89% |
| TrustRadius | 8.3% | −92.2% |
Number five, Software Advice (12.8% of citations), is missing from the table — the study doesn’t report a traffic change for it. And one caveat up front: the dataset skews heavily toward B2B software, which is why Gartner, G2, and Capterra dominate. Other segments will have different platforms and the effect may be weaker or stronger; there’s no particular reason to expect the pattern is unique to software, though.
AI doesn’t read your brand directly — it reads the middlemen
Our read on those numbers: a review platform used to be the place someone stopped by before buying. Now it’s increasingly not a person who stops by but a model — and it hands the conclusion over in an answer. The middleman between you and the customer didn’t disappear; it just moved from the browser into the answer.
In our article on brand mentions we wrote that AI assembles its picture of a company from other people’s sites. Here’s that in numbers: on evaluative queries you often aren’t competing primarily to get your own site cited — you’re competing over what’s written about you on a domain AI treats as the more credible source of ratings.
What changes in practice
Here’s the consequence the numbers point to, and the one most companies haven’t worked into how they operate:
- Traffic from the platform stops being the headline metric. A G2 or Trustpilot profile nobody clicks through from can still shape what AI says about you. Judging it by referral visits means measuring yesterday’s role.
- The state of the profile matters more than being there at all. The model reads whatever is sitting on the page: a half-empty profile with three reviews from 2023 conveys exactly that. Freshness and completeness aren’t cosmetic — they’re data.
- Your own site and third-party platforms play different roles. Your site is what you say about yourself — and how to set it up for reviews is covered in reviews and ratings for AI. Platforms are the corrective you don’t write. A strong position needs both, because models compare both.
A handful of platforms take most of the citations
Review-site citations aren’t spread evenly. The top 5 platforms collected 88 percent of all citations in the category; smaller sites like AlternativeTo or SaaSHub got almost nothing in the dataset.
A second SE Ranking study lines up with that — this one across 129,000 domains and ChatGPT citations. Domains with profiles on several review platforms (Trustpilot, G2, Capterra, Sitejabber, Yelp) averaged 4.6 to 6.3 citations; domains with no such presence averaged 1.8.
Practically: spreading yourself across ten directories doesn’t help. What helps is being genuinely present on the two or three that take the citations in your segment — and you find those by looking at who AI actually names on your queries.
They’re strongest on review and experience queries
It depends on the query type. Review platforms showed up in
- 49% of AI Overviews on queries explicitly about reviews (“X reviews”),
- 39.4% on general software and tools queries,
- only 17.1% on “best X” queries.
What follows from that: if customers come to you through review and experience queries, platforms are a priority. If your fight is mainly over “best X,” rankings and comparison articles matter more. So don’t treat platforms in isolation — treat them according to how customers actually ask about you.
What this means for your business
To be straight about it: both studies measure the US and global picture, weighted toward B2B software. Consumer and local platforms haven’t been measured this way. Three situations, three different conclusions:
- B2B or software: G2 and Capterra are a measured, documented channel. Given how concentrated the citations are, focus on those rather than the long tail of directories. The wider context is in AI SEO for B2B.
- Consumer e-commerce: the broad-audience platforms — Trustpilot, Sitejabber, Yelp — appear in the ChatGPT citation study, but nobody has measured their weight in AI answers the way the B2B sites were measured. Treat them as a reasonable bet to verify, not a documented certainty.
- Local services: your main field is Yelp and your Google Business Profile, plus the reviews on them; the global B2B platforms pass you by.
In all three cases, start the same way: write down the queries a customer would use to find you — “[brand] reviews,” “[product] experiences,” “best [category],” “[you] vs [competitor]” — and run them through AI tools. Whoever the answers name as a source is your platform list. The process is in the AI visibility test.
Common mistakes
Judging platforms by the traffic they send
Referral traffic is falling for everyone. The platform’s role moved to what it conveys about you to models.
Being on ten platforms instead of the right two
The top 5 sites took 88% of the category’s citations. Pick based on who AI cites on your queries.
Leaving the profile half-empty and unanswered
The model reads what’s sitting there. Three reviews from 2023 with no company reply is a message too.
Buying or faking reviews
It breaks platform rules — and vendor-written text is exactly what this mechanism is designed to route around.
What the data does not say
It doesn’t recommend a specific platform — that depends on your segment and on who AI actually names on your queries.
It doesn’t say anything measured about consumer or local platforms. The datasets are dominated by US B2B software; everything outside that is extrapolation.
It doesn’t say a profile guarantees a citation. Correlation isn’t a guarantee; a profile is a precondition you can work with.
Takeaways
Review platforms went through an odd transformation: they lost most of their people but kept their seat at the table where answers get assembled. Anyone who wrote them off based on the traffic curve is looking at the wrong chart.
A decision framework to close on: stop judging platforms by the visits they send you. Find out who AI cites on review and experience queries in your field. And pick one or two platforms that show up there and keep them alive — real reviews, a complete profile, substantive replies. It’s a plain tactic, but it’s the only one aimed at what models are looking for on those sites: text the vendor didn’t write.
Want to know who AI cites on queries in your field, and whether you’re among the sources? An AI SEO audit from Sniper Design walks your brand’s visibility across AI tools and the sources they draw on. And if you need to strengthen the evidence on your own site alongside third-party platforms, the page templates in the AI SEO Wireframe Pack cover that.
Sources: two SE Ranking studies — an analysis of review platforms in AI Overviews (30,000 keywords, 22,729 AI Overviews, snapshot 12/2025) and an analysis of ChatGPT citation factors (129,000 domains, 216,524 pages). Accurate as of July 19, 2026 — these numbers move fast, and we’ll keep the article updated.
