G2 and Capterra reviews now drive most AI citations for software recommendations. Here's the data on why ChatGPT recommends competitors, and how to fix it.
G2 and Capterra aren't winning because they write better content than your blog. They're winning because AI answer engines treat a review-site profile as close to a mandatory credential before they'll say a software brand's name out loud. Quoleady tested dozens of "alternatives" prompts on ChatGPT and found 99% of the tools it recommended had a G2 page, and 100% had a Capterra page, per Quoleady's LLMO research. That's not a stylistic preference. It's closer to a gate.
The fix isn't picking one over the other. It's owning both: the review-site infrastructure the engines check first, and the dated, citable content a review profile structurally cannot produce. Below is the data on why this happens, why ChatGPT sometimes recommends a worse product than yours, and what to actually do about it. This connects to the wider mechanics we cover in answer engine optimization, the primer for why AI citation behaves nothing like a Google ranking.
AI answer engines don't sample review sites lightly. They lean on them hardest exactly where a buying decision gets made, and the G2 ecosystem now owns the majority of that lean.
Presence, not persuasion, is the first filter. Quoleady ran dozens of SaaS "alternatives" keywords through ChatGPT and found that review-platform presence was "almost always required" for a tool to appear in the answer at all: 99% of recommended tools had a G2 page, 100% had a Capterra page, and 78.8% had a Wikipedia page, per Quoleady's LLMO research. If your product has no review-site footprint, you may simply never enter ChatGPT's candidate list, no matter how good your blog is.
Within the review-platform citation category specifically, one company now dominates. G2, together with Capterra, Software Advice, and GetApp, commands 84% of review-platform citations across roughly 35,000 ChatGPT citation URLs Profound captured in December 2025, versus Gartner at 13%, SourceForge at 2%, and TrustRadius at 1%, according to G2's own analysis. That concentration deepened on January 29, 2026, when G2 announced it would acquire Capterra, Software Advice, and GetApp from Gartner, a deal expected to close in Q1 2026 and reach a combined 200 million-plus annual software buyers with nearly six million verified reviews, per G2's own announcement.
The early citation-share modeling on that consolidation is stark. Pre-acquisition, G2 alone held 2.09% of all bottom-of-funnel AI citations, already the #4 domain overall; combined with the three acquired brands, the merged entity is projected at 3.68%, a 76% jump that moves it to #2 behind only Reddit, per Beomniscient's analysis. On proof-and-evidence queries specifically, the kind a buyer asks right before choosing, G2 was already #1 pre-merger at 7.54%, projected to 12.69% combined, a 93% lead over the next-closest source. Kevin Indig, a growth advisor to G2, frames the underlying shift bluntly: "Over 50% of decision-makers now initiate their software purchase journeys in LLMs like ChatGPT rather than traditional search," as quoted in the same Beomniscient piece. He's advising the company whose acquisition he's commenting on here, so read the number as directionally right rather than neutral.
Review sites don't just get cited, they get cited more as the buyer gets closer to a decision. Across the same ~35,000 ChatGPT citations, review platforms made up 8.6% of citations overall, but their share rose from 7.4% at the discovery stage to 13.2% at the evaluation stage, roughly 1.8x, per G2's funnel-stage breakdown. Your blog post explaining a category concept can win a discovery-stage query. The evaluation-stage query, the one where the buyer is deciding between you and a named competitor, is exactly where review-site weight climbs, and it's also exactly where a comparison query lands; we cover that query shape on its own in GEO for comparison pages.
Because ChatGPT isn't scoring product quality. It's checking whether a candidate clears a review-site presence threshold, then picking from whoever's left using signals that have little to do with how good the tool actually is.
This is the part that should be reassuring, and also the part that should worry you. Once Quoleady controlled for whether a tool appeared in review platforms at all, review score and volume showed almost no relationship to how high ChatGPT ranked it: Capterra review count correlated with ranking at -0.21, G2 review count at -0.16, G2 score at -0.11, and Capterra score at a negligible +0.02. Domain Rating was the strongest predictor they tested, per Quoleady's LLMO research. So a five-star G2 profile with 40 reviews doesn't automatically outrank your four-star competitor with 400. The gate is binary: show up on the review sites or don't. Past that gate, the model is pulling from a mix of off-site authority and whatever combination of sources it retrieved for that specific prompt, and it's mechanical, repeatable, and largely indifferent to your rating.
That mechanical, indifferent-to-rating behavior is also why ChatGPT gets recommendations visibly wrong so often. G2's own buyer survey found 64% of B2B buyers encounter inaccurate AI software recommendations frequently or very frequently, per G2's "Answer Economy" research. A worse product with a thicker review-site presence and stronger off-page authority can out-cite a better one that skipped the infrastructure.
Here's the part that makes this worth fixing today, not next quarter. When G2 asked buyers what made them trust an AI-generated recommendation, 45% named a software review-site citation as the single most confidence-inspiring signal in the answer, per the same G2 research. The AI answer isn't the trust mechanism. The review-site link inside it is.
That's also why the stakes are real: 69% of buyers say they picked a different vendor than they'd originally planned based on AI guidance, and 33% bought from a vendor they hadn't heard of before the chatbot named it. G2's Chief Innovation Officer Tim Sanders summarizes the shift this way: "Buyers have moved from reference to inference. Instead of gathering sources and synthesizing the data themselves, they trust AI chatbots to return the shortlist in a single prompt." That shortlist habit is spreading fast, too: 51% of B2B software buyers now start research with an AI chatbot more often than Google, up from 29% in April 2025, and 71% rely on AI chatbots for software research at all, up from 60% seven months earlier.
Don't build one review-site strategy and assume it covers every engine. ChatGPT and Perplexity pull from almost entirely different shelves of the web, and the review-platform weighting inside each shelf differs too.
Across a separate study of 456,570 citations spanning ChatGPT, Perplexity, Google AI Overviews, and Copilot between May 18 and June 18, 2025, G2 alone held 8.25% of ChatGPT's B2B software citations and 11.70% of Copilot's, out of every citation ChatGPT and Copilot produce for any topic, not just review platforms. Narrow the lens to just the review-platform slice of ChatGPT's citations (the same narrower category the 84% figure above measures), and GetApp specifically took 47.65% of that pool, per Hall's review-platform citation analysis. Different denominators, same story: whenever ChatGPT or Copilot cites a review site at all, it's almost always one wearing a G2 badge. ChatGPT and Copilot both concentrate their review-platform trust inside the G2 family, which lines up with what we cover on the ChatGPT side of the engine-by-engine playbook in how to get cited by ChatGPT, Perplexity, and Claude.
Perplexity doesn't play the same game. In the same 456,570-citation study, Perplexity leaned toward Gartner at 18.06% of its review-platform citations rather than the G2 ecosystem, and its citation habits skew away from directories generally: in Ahrefs' Brand Radar study of 3.1 million US queries (June 2026), YouTube took 32.4% of Perplexity's citations, Reddit 16.6%, and Wikipedia 8.2%, the most community-and-video-weighted mix of any assistant Ahrefs measured. And the two engines barely overlap on sources to begin with: only about 11% of the domains ChatGPT cites also turn up in Perplexity's citations, per Averi's analysis of 680 million AI citations. A review-site-heavy play optimized for ChatGPT does not transfer to Perplexity's answers; you're not competing on one shelf, you're stocking two.
None of this is exotic, but it is two separate jobs, and skipping either one leaves you invisible to a real slice of the buyers now starting their research inside a chatbot instead of Google.
Claim and fully build out your G2 and Capterra profiles the way you'd treat your pricing page: category, features, integrations, and a real review-generation motion, not a placeholder listing. Given that 99% of tools ChatGPT recommends have a G2 page and 100% have a Capterra page, a missing or thin profile is closer to a disqualification than a missed opportunity. Since rating and review count barely move the ranking once you're present, don't over-invest in chasing a perfect score; invest in simply being there, completely, on both platforms, plus Wikipedia where your company is notable enough to qualify.
A review-site listing is a static feature grid and a star average. It cannot publish a dated, sourced answer to "why does ChatGPT recommend my competitor," walk through a tested workflow, or cite a named authority with a direct quote, the exact three levers proven to raise AI citation odds. This is the owned-content half of the fix, and it's the half a directory can never take from you: answer real buyer questions in the first line under each heading, attach a verifiable, dated statistic to every claim, and cite named sources the way this post does. For the mechanics of getting that content extracted specifically by ChatGPT, see how to rank in ChatGPT.
Mentioned and cited are not the same thing, and the gap is bigger than most teams assume. ChatGPT mentions a brand about 3.2x more often than it links or cites it, 44% of prompts return zero brand mentions at all, and transactional, buying-stage queries trigger 8 to 10 times more brand mentions than informational ones, per BrightEdge's analysis of ChatGPT citation triggers. Unlinked mentions still matter for reputation. Across 75,000 brands, Ahrefs found unlinked branded web mentions correlate with AI visibility at roughly 0.66 to 0.71, far stronger than backlink metrics, which means being talked about, on review sites, in comparison posts, in your own blog, compounds even without a click. Run your own buyer-style prompts across engines weekly and log whether you're named with a link back to your domain or just named. Our AI citation tracking guide has the full setup, including the GA4 regex that catches the traffic your review-site citations eventually send.
Review-site presence and owned content aren't competing strategies, they're the same buyer's journey split into two moments. The review-site listing gets you into the shortlist. The dated, sourced blog post is what earns the citation once a buyer asks a follow-up question a static profile can't answer. Do only the first and you're one of ninety-nine similar-looking G2 listings. Do only the second and you may never clear the gate that gets you named at all.
Winning the G2/Capterra gate gets you into the conversation. Citable, dated content is what wins the follow-up question. Lyra writes answer-first posts with a verified stat per section and opens each one as a pull request you review and merge.
FAQ
Usually because your competitor has a G2 and Capterra profile and you don't, or yours is thin. Quoleady's study of ChatGPT's software 'alternatives' answers found 99% of recommended tools had a G2 page and 100% had a Capterra page. Review-platform presence functions as a near-mandatory inclusion filter before ChatGPT will name a tool at all, so a missing or sparse profile can knock you out of the answer regardless of product quality.
Barely. In Quoleady's data, G2 review count correlated with ChatGPT ranking at -0.16, G2 score at -0.11, and Capterra score at just +0.02, all close to zero. Domain Rating was the strongest predictor they tested. Once you clear the presence bar, more reviews or a higher score does very little to move you up the list.
No. In a 456,570-citation study spanning ChatGPT, Perplexity, Google AI Overviews, and Copilot, GetApp alone took 47.65% of ChatGPT's review-platform citations while Perplexity leaned toward Gartner at 18.06% instead. Perplexity's overall citation mix skews further still, toward YouTube and Reddit over any directory. A strategy built only around the G2 ecosystem transfers poorly to Perplexity.
Run your own buyer-style prompts across ChatGPT, Perplexity, and Copilot weekly and log whether a link back to your domain appears next to your name, versus your name simply being spoken with no source attached. ChatGPT mentions brands about 3.2x more often than it cites them, so a mention with no citation is common and worth tracking separately.
Built by the tool you're reading about
Lyra finds the topics worth ranking for, writes them in your repo's voice, fact-checks every claim, and opens a pull request scored and ready to merge. You review and hit merge. Want to see what she'd write for you? Start free with three posts, no card.
Keep reading
AI writing tool data training policies vary sharply by vendor: the exact contract language to check before your SaaS drafts train someone else's model.
AI writing tool security review checklist: what SOC 2 Type II, SSO, data residency, and audit logs an enterprise procurement team asks before they sign.
Reddit is in roughly 21% of Google AI Overviews and 44% of their social citations. Here's why, and how your blog can borrow its structure to earn citations too.