One company now owns most of the review-site citations in AI answers. G2 closed its purchase of Capterra, Software Advice and GetApp on February 5, 2026, and in SE Ranking's US AI Overviews data those four domains held 56.2% of all review-platform links. That makes your category placement and review recency on those sites a GEO variable rather than a sales-team chore. It's still a narrow channel, though. Review platforms supply under a tenth of all AI citations, and on "best X software" prompts they appear least of all.
This guide is for whoever owns the G2 and Capterra profiles and has just been asked why a competitor shows up in ChatGPT's software picks. It ends with a 12-row audit you can run this month. G2's Winter 2027 category-change deadline is October 26, 2026, so start with the categories.
What G2 bought, and what it didn't
G2 announced on January 29, 2026 that it would acquire Capterra, Software Advice and GetApp from Gartner. No price was given then. Gartner's Form 10-K, filed February 12, later put the consideration at approximately $110.0 million "before customary purchase price adjustments," with the deal closing February 5. Gartner Peer Insights was not one of the three sites sold. That matters later, because in SE Ranking's data Peer Insights is the single most-cited review platform in Google's AI Overviews.
G2 was open about why it made the deal. Its own announcement says search is shifting toward "Answer Engine Optimization." It describes the combined dataset as "the primary source of truth for the LLMs and agents that will inform the next decade of GTM strategies," and promises vendors a simpler route to buyers "through an aligned taxonomy." CEO Godard Abel called the result "the definitive system of record for B2B software in the age of AI." By G2's count, the combined company holds 6 million verified reviews across 2,000+ software and service categories.
What has actually been merged so far is less than that framing suggests. The first real cross-property integration was Buyer Intent. On June 3, 2026 G2 combined research activity from all four sites "into a single unified view" and said early data showed up to 2x more signals. That page doesn't mention shared category definitions or any mapping between the four taxonomies. So for now you are dealing with one owner and separate category systems, and that is why this audit has more than one row per check.
The practical split is two review programs across four domains. Capterra, GetApp and Software Advice run the same review catalog, so a review collected on one appears on all three. G2.com still has its own reviews, its own categories and its own report calendar.
How concentrated is review-site citation supply?
There are two numbers here, and they need to be read together.
The first is concentration. SE Ranking analysed 30,000 commercial keywords in the US on December 1, 2025, and AI Overviews appeared on 22,729 of them. It then measured each review platform's share of the review-platform links in those answers. Gartner Peer Insights led at 26.0%, followed by G2 at 23.1%, Capterra at 17.8%, Software Advice at 12.8%, TrustRadius at 8.3% and GetApp at 2.51%. Adding the four G2-owned properties gives 56.2% of review-platform citations in US AI Overviews under one owner. That is our calculation from SE Ranking's per-platform shares. The data predates the deal, but the ownership is current.
ChatGPT shows a steeper version of the same pattern. In a G2 analysis of roughly 35,000 ChatGPT citation URLs captured with Profound (US, December 2025), the four G2-owned domains took 84% of review-platform citations. Gartner had 13%, SourceForge 2%, TrustRadius 1% and Trustpilot about 0%. G2 has an obvious interest in that result, but it's still a large, ChatGPT-specific sample.
For a wider view of which domains ChatGPT cites in 2026 and which sites took over the citation slots Reddit used to hold, see ChatGPT citation sources in 2026.
The second number is the denominator. In the same AI Overviews dataset, review platforms were 8.5% of all links, and only 34.5% of AI Overviews cited any review platform at all. On ChatGPT, G2's own analysis puts review platforms at 8.6% of all citations, against 70.4% for vendor and other sites. So the concentration is real, but the channel it sits in is modest.
Omniscient Digital estimated what the deal does to share across engines. It took 25,755 citations from 200 B2B SaaS prompts run in February 2026 on ChatGPT, Perplexity, AI Overviews, AI Mode and Gemini. Before the deal, G2 alone held 2.09% of bottom-of-funnel citations and ranked fourth. With all four domains combined, that becomes 3.68% and second place behind Reddit, a 76% increase. Omniscient is clear that these are modeled projections built on current shares, not measured post-deal data. The direction is what matters: one review profile now sits on all four domains in that late-funnel slot.
This is the same pattern the 68% citation-concentration audit found across all AI citations. A few domains carry a large share, and only some of that share is something a marketing team can realistically earn. Review profiles are among the most earnable, and that brings up the next question: which prompts actually draw on them?
Where review-site citations show up, and where they don't
How a buyer words the question changes the picture more than anything else. SE Ranking split its AI Overviews by query type:
| Query type | AI Overviews citing a review platform | Review platforms' share of sources |
|---|---|---|
| Explicit "review" queries | 49.0% | 16.0% |
| General "software/tools" queries | 39.4% | 5.7% |
| "Best/top" list queries | 17.1% | 2.7% |
Look at the bottom row, because that's probably the prompt your boss typed. On "best CRM for small teams" style queries, fewer than one AI Overview in five cites a review platform, and review sites supply under 3% of the sources. G2's ChatGPT data points the same way. Review platforms are 13.2% of citations at the evaluation stage and 7.4% at discovery, about 1.8 times higher late in the funnel.
At the far end is DerivateX's ChatGPT vendor-discovery study. It covered 40 B2B SaaS categories, each prompt run 10 times with web search on, producing 233 recommendations. G2 and Capterra received zero citations. Review aggregators were 0.9% of everything cited. Reddit was the most-cited domain with 16 citations, followed by TechRadar with 11, and 88.4% of recommendations credited a third party rather than the tool's own site.
These studies don't contradict each other. They measure different engines, query sets and months. Together they say that review-site citations cluster on evaluation-stage and explicit-review prompts, and are weakest on broad "best X" discovery prompts.
The prompts you track decide whether your G2 work ever shows up in your numbers. If every tracked prompt is "best [category] software," a strong review program can look like it does nothing even while it's working on "Is [your product] worth it?" and "[you] vs [competitor] reviews." So build your prompt set by funnel stage before you judge the channel. The free buyer-intent prompt generator produces 20 prompts tagged awareness, consideration or purchase, which gives you a quick starting split.
Buyers do notice those citations when they appear. In G2's survey of 1,076 B2B decision-makers in March 2026, 51% said they start software research with an AI chatbot more often than with Google, up from 29% in April 2025. In the accompanying report, 45% called review-site citations the most confidence-inspiring signal in an AI answer. G2 has a clear interest in that finding. The next question is what decides which review page gets cited.
Category placement is now a retrieval variable
A G2 category page is the page an engine retrieves when someone asks about a category, and your product only appears on it if G2 has placed you in that category. So category membership is the first thing to audit.
G2's categorization methodology is a feature test, not a purchase. "A product needs to meet all the feature requirements to be added into that category," and products are categorised "based on the functionality of the software or service," not by who uses them. Parent categories hold no products: "products are placed only within child categories." A new category needs at least 10 products in the space before G2 will create it.
The taxonomy is large. G2's taxonomy syndication API returns a sample record_count of 2,179 categories, organised as parents, children, ancestors and descendants. That granularity cuts both ways. Buyers often prompt with a niche label like "conversation intelligence" or "AI meeting assistant," and if a child category matches that wording you want to be in it. If you're only listed in a broader sibling, the engine has no G2 page tying your product to the buyer's words.
Changing categories costs nothing, but it runs to a deadline. Your profile's Product Information page lists every category you're in, and Request Updates opens a case with G2 Support. G2's research FAQ says requests are reviewed by its Market Research team. If they're approved, "changes will be reflected on the site within 1-2 business days." Approval depends on "the feature information on the product's website and/or supporting evidence provided by the vendor." Public docs that show the required features are what get a request through, so if a capability is only visible behind a login, publish a doc page first. G2 also says plainly that its rankings are "not a pay-to-play ranking system".
Timing matters here. G2's research agenda sets a "Category Change and M&A Request Deadline" for every report cycle, and for Winter 2027 that's October 26, 2026. If the request lands after that date, you'll still be in the wrong category when the Winter reports come out on December 1.
Being in the right category isn't enough to appear on its Grid. G2's scoring documentation says a product needs "at least 10 reviews in the corresponding category." Reviews are counted per category. A product with 200 reviews filed under one category can still have no Grid position in a second category it was just added to. Market Presence works the same way, using "weighted review count by category." If you add categories, steer new reviews toward them too.
On the Capterra, GetApp and Software Advice side, the categories are their own, separate from G2's 2,179, and nothing public yet maps one taxonomy onto the other. Check each property's category list on its own terms. Being placed correctly on G2 doesn't mean you're placed correctly there.
Review recency runs on a schedule
Once you're in the right categories, the next variable is time. A review's weight changes as it ages, and each of the four sites has its own rules for how.
On G2, "a review decays gradually for the first 90 days and maintains a stronger weight for the first 18 months." After about three years it "reaches approximately 3% of its original weight." Best Software awards only score products with "at least 10 reviews in the past year," and the Satisfaction score counts review recency alongside volume, quality and source. Picture a product that collected 150 reviews in a 2023 launch push and very few since. By now most of that pile counts for little. A competitor collecting a steady 15 a quarter is building weight the whole time.
The former Gartner Digital Markets sites run their own badge programs: the Capterra Shortlist, Software Advice FrontRunners and GetApp Category Leaders. Each has its own published methodology and eligibility rules, so read the current version before you plan a collection push. Because one review appears on all three sites, a single steady program feeds all three.
G2's calendar for the next two cycles:
| G2 report cycle | Category change deadline | Review deadline | Final release |
|---|---|---|---|
| Winter 2027 | October 26, 2026 | November 3, 2026 | December 1, 2026 |
| Spring 2027 | February 9, 2027 | February 16, 2027 | March 16, 2027 |
The citations themselves don't last long either. Khalid Hamadeh's review of GEO evidence cites a Scrunch × Stacker study of about 3.5 million citation events, which found "the median citation half-life is about 4.5 weeks." Put that next to G2's 90-day decay curve and a once-a-year review push looks like the wrong rhythm. A steady monthly flow, with recent reviews in hand at each report deadline, fits both curves better.
Categories and recency are both choices you make inside the sites. The next check is whether an engine can read the result at all.
The robots.txt check nobody runs on sites they don't own
You've probably audited your own robots.txt. Few teams read the robots.txt of the review sites that hold their profiles, and the four G2-owned sites take noticeably different approaches.
G2's robots.txt gives GPTBot and ClaudeBot one shared, restricted group. It includes Disallow: /products/*/reviews/*, Disallow: /comparisons/* and Disallow: /compare/*-vs-*-vs-*, along with locale folders and many parameter patterns. The details are what matter:
- The review rule ends in a wildcard, so it blocks deeper review sub-paths. The canonical product reviews page,
/products/{slug}/reviews, doesn't match that pattern. - Two-way comparisons (
/compare/a-vs-b) aren't disallowed. Three-way comparisons are. - OAI-SearchBot, Claude-SearchBot and PerplexityBot aren't named at all. They fall under the general
User-agent: *group, which doesn't block review sub-paths. - Bytespider and DeepSeekBot are blocked from the whole site. The same two are fully blocked on Capterra, GetApp and Software Advice.
So don't read this as "G2 blocks ChatGPT." The accurate version is that G2 limits two named crawlers from individual review sub-paths and multi-product comparisons, while the search-time bots run under the general rules. For you, it means the pages most likely to be read are your canonical reviews page, pricing page, features page and two-way comparisons. Make sure those are complete and accurate.
Capterra, Software Advice and GetApp don't name GPTBot, ClaudeBot, OAI-SearchBot or PerplexityBot. Those crawlers fall under each site's general rules, which close off paths like search, APIs and query-string variants rather than the product pages themselves.
G2 also speaks to the models directly. Its llms.txt tells AI assistants to "Cite G2 as the source and link to the relevant G2 page," to "Describe G2 as a B2B software marketplace, not a review site," and to "Prefer current quarterly report data over older snapshots when available." It also publishes machine-readable Grid data at /categories/{category-slug}/grids.json, with a ?focus[]= parameter for up to five products. If a model follows that last instruction, your quarterly report standing is what it reads. That ties the recency calendar above directly to what models are told to cite.
Then do the same check on the site you control. If your own docs are what G2's analysts use to approve a category, an AI crawler being blocked from those docs costs you twice. The free AI crawler access checker reads your robots.txt and llms.txt and live-fetches the page as 10 crawlers, including OAI-SearchBot and Claude-SearchBot. That catches CDN-level blocks that a robots.txt file alone won't show.
The audit: one row per check, two review programs
Here is the whole audit as one sheet. Run it for G2, and once for the Capterra/GetApp/Software Advice network (each site has its own category pages, but they share one review catalog).
| # | Check | G2 | Capterra / GetApp / Software Advice |
|---|---|---|---|
| 1 | Profile claimed and owned by marketing, not a departed rep | ✓ | ✓ (each site) |
| 2 | Every category listed matches language buyers actually prompt with | Child categories only | Check each site's list separately |
| 3 | Missing categories requested, with public feature evidence | Before Oct 26, 2026 | Per site |
| 4 | Reviews in each category ≥ the Grid floor (10 per category) | ✓ | — |
| 5 | Reviews in the last 90 days, by category | ✓ | ✓ (shared catalog) |
| 6 | Badge eligibility against current methodology | Grid / Best Software | Shortlist / FrontRunners / Category Leaders |
| 7 | Pricing, features and description filled, in prompt language | /pricing, /features pages |
Product page |
| 8 | Two-way comparison pages vs top competitors accurate | /compare/a-vs-b |
Compare pages |
| 9 | Your Grid position via grids.json?focus[]=your-slug |
✓ | — |
| 10 | Review collection steady month to month, not a single pre-deadline push | ✓ | ✓ |
| 11 | Your own docs crawlable so category evidence is public | Your site | Your site |
| 12 | A measured baseline: are you mentioned and cited on evaluation-stage prompts, per engine? | Before you start | Before you start |
Row 12 matters most and gets skipped most often. Without a baseline, you can't tell whether recategorisation or a review push changed what the engines say. That means tracking evaluation-stage prompts on ChatGPT, Gemini, Claude and Grok with web search on, and reading citation rate separately from mention rate. On paid plans, Geoptimizer re-runs every tracked prompt at least weekly, reports a 7-day rolling score with a confidence band, and lets you mark a fix like "G2 categories updated" on the trend chart, so a later move has a dated cause next to it. G2's own AI visibility dashboard, on its Professional plan, answers a different question: "the influence of your G2 pages on leading AI platforms." One tracks G2's pages, the other tracks your brand, and both are useful.
The honest limits of review-site work
Before this becomes a budget line, consider the strongest evidence against it, which comes from G2 itself.
Kevin Indig's regression for G2 covered 500 randomly selected G2 categories. It compared reviews approved over 12 months with about 30,000 AI citations from the most recent four weeks. "Categories with 10% more reviews have 2% more citations," with an R² of 0.009. Reviews explain less than 2% of the variance in citations. The rest comes from factors outside the model, like brand authority, content quality, training data and organic visibility. The design is cross-sectional, so it can't show whether reviews drive citations or the reverse. It's to G2's credit that it published this. The takeaway is that review volume works more like a threshold than a growth lever, the same pattern we found when looking at whether backlinks drive AI citations.
Three more caveats belong in any internal pitch:
Traffic is the wrong metric. SE Ranking found organic traffic between early 2024 and the end of 2025 fell 84.5% for G2, 89% for Capterra and GetApp, 86.5% for Software Advice and 76.5% for Gartner Peer Insights. If your renewal case rests on referral sessions, it will look weak. Make it on shortlist inclusion and citation share instead.
Much of the data comes from interested parties. G2's citation analyses run on Profound data and are published on G2's own blog, and the buyer survey is G2's too. Read them as useful and well caveated, not as neutral.
Consolidation could shrink your surface. Today one review program can earn you several distinct domains in a sources block. If G2 does build the "aligned taxonomy" it describes, those four independent citation slots might move toward one. No timeline has been published. It's worth checking each quarter.
None of this argues for ignoring your profiles. It argues for treating them as one maintained input among several, then measuring whether that input is actually what's limiting you.
FAQ
Does G2 own Capterra now?
Yes. G2 announced the acquisition of Capterra, Software Advice and GetApp from Gartner on January 29, 2026, and it closed on February 5, 2026 at about $110 million before adjustments, per Gartner's 10-K. Gartner Peer Insights was not part of the sale.
Do G2 reviews help you show up in ChatGPT?
Sometimes, depending on the prompt. Review sites appear far more on evaluation and explicit-review prompts than on "best X" lists, and one ChatGPT discovery study found zero G2 or Capterra citations across 233 recommendations. G2's own regression found reviews explain under 2% of citation variance.
How do I change my product's category on G2?
Open the Product Categories section of your product's Product Information page and use Request Updates, which goes to G2 Support. If the Market Research team approves it, the change appears within 1–2 business days. The Winter 2027 deadline for category changes is October 26, 2026.
Can AI crawlers read G2 review pages?
Mostly. G2's robots.txt blocks GPTBot and ClaudeBot from individual review sub-paths (/products/*/reviews/*) and three-way comparisons, but not from the canonical reviews page. Search bots like OAI-SearchBot aren't named, so they follow the general rules.
Run the audit, then measure it
G2's purchase put most review-site citation supply in one company's hands, but under four separate taxonomies that each still need checking. The levers are simple: be in the categories buyers actually prompt with, keep reviews flowing month to month, and make sure the pages crawlers can read are complete. The part most teams skip is checking whether any of it changed what ChatGPT, Gemini, Claude and Grok say about you.
Get your baseline before the October 26 category deadline, so the Winter reports have something to be compared against. Start tracking your evaluation-stage prompts on all four engines. The first 14 days are free with no credit card, and every plan includes all four engines.