G2 is not driving AI visibility for most software categories. This is what is.

The common assumption is that G2 and Capterra are the gatekeepers for software AI visibility -- get your reviews there, show up in AI answers. We ran live AI prompt panels across 270 software brands in three categories and checked which domains actually appeared in the citation sources. G2 did not crack the top five in two of the three categories.

The data and how we collected it

We ran 10-prompt buyer-intent panels for each of 270 software brands across three categories: customer support software (58 brands), project management software (173 brands), and HR software (39 brands). Each panel used four AI engines that pull live web search results: ChatGPT, Gemini, Perplexity, and Claude. The prompts were framed as buyer questions -- "what are the best options for customer support software for a 50-person company" and similar -- so the citation sources reflect what engines reach for when advising a real buyer.

We excluded Gemini's citations from the source analysis. Gemini routes 99.8% of its citations through a single proxy domain (vertexaisearch.cloud.google.com) rather than reporting the underlying page URLs, so it contributes nothing useful to a per-domain breakdown. The counts below are from ChatGPT, Perplexity, and Claude only.

Customer support software: editorial sites, then review platforms

Across 58 brand panels in the customer support category, the top citation sources were a mix of CX-focused editorial sites, competitor product pages, and tech review publications. G2 appeared -- but at rank six.

The top five sources together account for 18% of all citations. The distribution is relatively diffuse -- no single domain overwhelms the others, and the engines pull from a mix of editorial blogs, competitor pages, integration-hub content, and review platforms.

Project management software: niche content sites dominate

Project management had the largest sample -- 173 brand panels. The citation source picture was more concentrated, and G2 did not appear in the top ten at all. The gatekeepers here are category-specific editorial sites and competitor product pages.

The Digital Project Manager alone accounts for 8.1% of all citations in the PM category -- more than triple what G2 gets in the customer support category. It appears because it publishes comparison articles, tool lists, and PM methodology content that AI engines treat as authoritative sources for buyer-intent queries in this space.

HR software: G2 finally cracks the top five -- at rank five

Across 39 HR software brand panels, the citation pattern looked like a hybrid: HR-focused editorial sites lead, with tech review publications and competitor pages in the middle, and G2 appearing at rank five.

G2 is present and meaningful here, but it shares influence with editorial sites and a major competitor's domain. A brand that focused only on G2 reviews and ignored peoplemanagingpeople.com or technologyadvice.com would be missing a larger share of the actual citation traffic.

What local services looks like for comparison

We ran a separate full-panel audit for a regional internet provider -- a category where comparison directories are purpose-built to index local options by geography. The citation structure is noticeably more concentrated, and the source type is different.

The top five account for 34.5% of all non-Gemini citations -- roughly double the concentration of the software categories. Three of the five are ISP-specific comparison directories that exist specifically to help consumers find local providers by zip code. A brand not listed on those three sites is absent from more than a fifth of all citations.

Why the patterns differ

Local service categories have geo-indexed aggregators. Broadbandsearch.net, allconnect.com, and similar sites are built around one task: given a zip code, return a list of ISPs. Because that is exactly what buyers are asking ("what internet providers are near me"), AI engines treat those directories as authoritative and cite them frequently.

Software categories have no equivalent. There is no geo-indexed database of customer support tools because buyers are not searching by location. Instead, the engines draw from wherever category knowledge has accumulated: editorial sites that publish tool comparisons (thecxlead, thedigitalprojectmanager, peoplemanagingpeople), tech review publications that cover software, integration hubs that document which tools connect, and the tools themselves through their product content.

Review platforms like G2 are part of this mix -- but they are not the dominant source in most categories. A G2 profile helps. It is not the gate.

What this means for the fix

The barrier to appearing on a local directory and the barrier to appearing on an editorial content site are very different.

For local services: creating or claiming a listing on broadbandsearch.net or allconnect.com takes a few hours and is often free. The fix is structural and fast.

For software: getting covered on thedigitalprojectmanager.com or thecxlead.com requires either appearing in their comparison roundups (which requires outreach, positioning, and often existing product reviews elsewhere) or building the kind of category authority that leads editorial teams to include you without being asked. That is a longer path.

The citation source list from an AI visibility audit tells you precisely where to focus. In some categories that is a directory listing. In others it is an editorial relationship with a category-specific site. In most software categories, it is both -- and they require different timelines and different tactics.

Methodology note

Panels ran across four AI engines (ChatGPT with web search, Gemini with grounding, Perplexity Sonar, Claude with web search). Gemini citation data is excluded from source analysis because Gemini proxies all citations through vertexaisearch.cloud.google.com, making per-domain breakdown impossible. Source counts reflect ChatGPT, Perplexity, and Claude only. Panels ran between June 2026.