Review Sites & AI Recommendations: G2, Capterra, Trustpilot

By Cory Maki · Oct 1, 2026

Ask ChatGPT which CRM to buy, ask Perplexity for Notion alternatives, or ask Google's AI Overview whether a SaaS tool is worth the money, and the answer is quietly reading review sites. Not your homepage. Not your feature page. G2, Capterra, Software Advice, TrustRadius, and Trustpilot sit between your brand and the recommendation — and most founders have never thought about them as a GEO surface.

This post is not another case study tracing one query's sources. I've already done that in How AI Picks Local Businesses, which proved review aggregators outrank brand sites. The question here is structural: why do these specific platforms win AI citations, how do their schema, index authority, and category pages feed retrieval, and what does the February 2026 G2–Capterra consolidation mean for where you spend effort? Review sites are a third-party consensus layer that AI prefers by design — the same mechanism that makes AI cite Reddit over your content, applied to structured commercial data.

Review sites are the second-most-cited source type in AI answers

Review and trust platforms are not a niche signal. In Seer Interactive's analysis of over 800,000 AI responses, review and trust websites are now the second most cited source type, accounting for 14% of all citations in AI responses. For software queries specifically, the share is higher. One SE Ranking study of 30,000 commercial searches found that despite traffic declines, review platforms like Gartner, G2, and Capterra remain heavily cited in AI Overviews for commercial queries.

The concentration is extreme. SE Ranking's analysis of review platforms in AI Overviews — 30,000 commercial keywords and 211,000+ cited links — found that five platforms account for 88% of all review-platform citations in Google AI Overviews: Gartner Peer Insights (26.0%), G2 (23.1%), Capterra (17.8%), Software Advice (12.8%), and TrustRadius (8.3%). That means a handful of domains intermediate nearly every software recommendation an AI produces.

This parallels why Reddit dominates AI search citations: engines prefer third-party aggregation over first-party marketing. Reddit supplies unstructured community consensus; review sites supply structured consensus — ratings, counts, categories, and feature tags that a retrieval system can parse without guessing.

The inclusion gate: presence is close to mandatory before AI will name you

The most important finding is not about citation share — it's about eligibility. Review-site presence functions as a trust gate before AI will put your product in a recommendation at all. In a 2026 study of B2B SaaS recommendations, 100% of the tools ChatGPT named had reviews on Capterra and 99% had reviews on G2. A tool with no presence on either almost never showed up.

The Trustpilot data shows the same cliff for consumer and services brands. In Seer Interactive's study, brands with no Trustpilot profile have a median AI citation rate of 1%. Brands with even a minimal profile — as few as 1–13 reviews — jump to 53.5%. That's a 52 percentage point swing. The implication is blunt: you don't need hundreds of reviews to show up. You just need to exist on the platform.

This is why absence is a reputation problem, not just a visibility one. AI systems do not only use Trustpilot when reviews are present. In some cases, they also mention Trustpilot when independent review evidence is missing — meaning a missing profile can itself become something AI flags. If you want to audit whether this is happening to your brand, start with the free AI visibility report.

Why these platforms win: structured consensus engines can trust

Three structural properties explain why review sites beat your own site in AI retrieval.

They solve the verification problem. LLMs cannot verify your marketing claims, so they outsource trust. Kevin Indig framed it precisely in his 30,000-citation study: "G2 provides three attributes that matter: verified buyers (reduces noise), standardized schema (machine-readable), and review velocity (current market activity)." LLMs face a verification problem and G2's structure solves it at scale.

They publish machine-readable, timestamped data. Trustpilot's edge is the same mechanic. It combines a high-authority domain, a huge volume of structured reviews, and timestamped, moderated data that engines find easy to trust and reuse. As Trustpilot's CEO put it, private signals like NPS and CSAT are invisible to models — public reviews on third-party platforms are readable, structured, and timestamped.

Their category pages are built for the exact query shape. A page titled "Best CRM Software" with ranked products, ratings, and filters is a near-perfect match for "what's the best CRM." G2 category ranking pages (e.g., 'Best CRM Software — G2') are among the most-cited pages in AI responses. This is the same passage-extraction logic I cover in content chunking for AI retrieval — review sites happen to pre-structure the perfect chunk.

Important nuance: review count is a weak lever past the inclusion threshold. The relationship is real but modest. The relationship is statistically reliable but modest. Reviews explain less than 2% of the variance in citation count. The remaining 98% is brand authority, content quality, organic search visibility, and model training data. In practice, asked for Notion alternatives, ChatGPT ranked Coda at number three with 97 Capterra reviews, while ClickUp, sitting on roughly 4,490 reviews, landed at number four. A 46x difference in review count did not translate into a better AI position. Clear the threshold, then stop chasing raw volume.

The platform map: G2 vs Capterra vs Trustpilot (and who owns them now)

The three platforms serve different query pools. The February 2026 consolidation changed the competitive map underneath them: G2 acquired Capterra, Software Advice, and GetApp from Gartner in February 2026, consolidating most of the review-platform citation category under one company. Mapping that onto the citation data: G2 now owns three of the five most-cited review platforms — G2, Capterra, and Software Advice — together 53.7% of review-platform citations in AI Overviews. Gartner kept only Gartner Peer Insights; TrustRadius stays independent. Even under one owner, the properties feed AI separately: despite common ownership, G2 and Capterra retain distinct audiences, category structures, and datasets that feed AI separately.

Platform	Best for	Who owns it	What AI pulls	Priority action
G2	Broad B2B SaaS, mid-market & enterprise; the default for most software	G2 (self)	Category Grid pages + product pages	Complete feature tags; hit the Grid review floor per category
Capterra	SMB software discovery; broad category coverage	G2 (acq. 2026)	Category listing pages; curated pros/cons block	Fix categories, verify pricing, qualify for the Shortlist
Trustpilot	B2C, services, and DTC; brand-level trust	Trustpilot (self)	TrustScore, review count, synthesized review themes	Claim the profile, respond to reviews, add Product Review Pages
Gartner Peer Insights	Enterprise software buyers	Gartner	Analyst-adjacent enterprise validation	Prioritize if your buyers are enterprise
TrustRadius	Enterprise, comparison depth	Independent	Product comparison pages	Secondary for enterprise B2B

The routing rule is simple: enterprise buyers trigger Gartner Peer Insights and TrustRadius more frequently; SMB and discovery queries favor Capterra and Software Advice. G2 spans the widest range of B2B SaaS queries and is a strong default priority for most software brands. For a plain positioning: G2 is the most cited for B2B SaaS, Capterra for SMEs, and Trustpilot for B2C.

The thresholds that actually gate G2 visibility are concrete. Appearing in a category Grid report requires at least 10 reviews in that category; the category qualifies for a Grid report with 6+ products at 10+ reviews each and 150+ reviews total. On Capterra, qualifying for the Shortlist requires at least 20 unique reviews in a trailing 24-month window.

The twist: review sites gate recommendations but rarely get the link

Here's the counterintuitive part that reframes strategy. Review sites decide whether you get recommended far more than they get the citation for it. G2 and Capterra almost never appear as the cited source in AI answers, even when the tools they list get recommended. In a 2026 study of 233 software recommendations, review aggregators accounted for just 0.9% of all citations, and G2 and Capterra each received exactly zero. AI read the category, named the tools, and then linked to something else entirely.

Two lessons follow. First, treat review-site presence as eligibility infrastructure, not a link-building play — it gets you into the consideration set, after which other sources carry the visible citation. Second, the measurable signal to track is your inclusion in the recommendation, not your profile's link count. That's a citation share question, not a referral-traffic one.

Trustpilot is the partial exception on the services/B2C side, where it both gates and gets cited. 45.2% of AI responses citing Trustpilot attributed content directly to the Trustpilot site, and when cited, a large share of ChatGPT answers referenced the TrustScore or star rating directly. That reflects Trustpilot's domain weight: according to Promptwatch, Trustpilot is the fifth most cited domain globally on ChatGPT — sitting above most software-specific review sites in AI's source hierarchy.

The index mechanic: why Bing and Google matter more than the AI engine

The reason review-site schema matters is where it lands, not whether an LLM parses JSON-LD at read time. Most engines don't. AI engines process schema markup indirectly, through search engine index enrichment, not by parsing JSON-LD semantically in real time. The mechanism: schema → search engine index enrichment → AI grounding. Your schema feeds Google's Knowledge Graph and Bing's entity index.

That's why the single confirmed schema signal runs through Bing. The only major AI platform to officially confirm schema usage is Microsoft Bing. Fabrice Canel, Principal Product Manager at Microsoft Bing, stated at SMX Munich in March 2025 that "Schema markup helps Microsoft's LLMs understand content." And since ChatGPT search relies on the Bing index, this confirmation extends further than it first appears. If you've read why ChatGPT can't see your website, this is the same pipeline: Bing index in, ChatGPT citation out. Review platforms are deep in Bing's index with authority most brand sites can't match — see how Microsoft Copilot picks sources vs ChatGPT.

Be realistic about what schema does on your own domain. Schema resolves ambiguity and reduces extraction friction. It does not compensate for absent authority. A page with clean, complete JSON-LD but no external mentions, weak topical coverage, and no inbound links is still a low-authority page. I cover which types actually get cited in Schema Markup for GEO.

What founders should actually do

The play is narrow and high-ROI. Review-site presence is one of the few GEO levers you fully control.

1. Clear the inclusion gate on the two platforms your category is cited on. Use the citation data: broad B2B SaaS → G2 first; SMB → Capterra; B2C/services → Trustpilot. Hit the Grid/Shortlist review floors, then stop chasing volume for its own sake.

2. Fill every structured field — the fields are the citation. Where people go wrong: treating structured fields as an afterthought while polishing the prose above them. It is backwards. The fields are the citation. Use feature tags that mirror competitor phrasing so an engine matching a buyer's query can match yours.

3. Keep your facts identical across every surface. If your G2 profile describes your product differently than your homepage does, you weaken the consensus signal. This is the cross-source consistency requirement applied to review sites.

4. Mirror review authority on your own domain with honest AggregateRating. Only include aggregateRating if you have real, verifiable review data from a credible third-party source (G2, Capterra, your own reviews API). Fabricated ratings torch your schema trust signals and invite a manual action. Point your Organization schema's sameAs at your review profiles — connecting entities with schema is how you tie the pieces into one entity.

5. Respond to reviews. On the trust layer this is measurable: brands with 95%+ response rates to reviews maintain citation share that low-response-rate competitors lose.

Review sites are the structured half of the third-party consensus layer; Reddit is the unstructured half (see the Reddit playbook for ChatGPT). Both exist because AI trusts outside corroboration over your own claims. Treat them as infrastructure, verify what AI currently says about you with the free visibility report, and fit this into the broader 90-day GEO roadmap. For the full framework, start with the GEO guide.