Open ChatGPT and ask it for the best CRM for a small sales team, or the best HR software for an Indian startup. It gives you a shortlist of names. That shortlist is not an opinion the model holds. It is assembled from a specific set of sources the model reads at answer time, and if you know which sources those are, getting into the shortlist stops being a mystery.
We classified every source ChatGPT cited across the B2B software categories in the August 2026 Cited Index, 379 citations in Conversational AI, CRM, and HR & Payroll. The mix is the whole strategy:
| Source type | Share of ChatGPT's B2B citations |
|---|---|
| Brand websites | 72% |
| Independent publishers (TechRadar and similar) | 15% |
| Community (mostly Reddit) | 7% |
| Review platforms (G2, Capterra) | ~1% |
| Marketplaces | 0% |
| Other | 4% |
Read the top and the bottom of that table together, because they are the two that should change where a SaaS team spends.
The 72%: ChatGPT reads what brands publish about themselves
For B2B software, ChatGPT is a documentation reader. Nearly three-quarters of what it cites is on brands' own domains: product pages, comparison pages, docs, pricing. This is the opposite of consumer categories, where a brand's own site is a minority source and the answer is assembled off-site. In software, the citation lives on a brand website, which means it can live on yours.
That is genuinely good news, because it is the one layer you fully control. The catch is that "a brand website" does not mean your website by default. In any given answer those 72% of citations are spread across every competitor in the category, so what decides it is whether your pages are the ones extractable enough to be picked. Thin marketing copy that says "the leading CRM for modern teams" gives ChatGPT nothing to lift. A comparison page that lays out who your product is for, what it costs, and how it differs, in plain extractable language, gives it a citation-ready answer.
The 1%: the review sites you optimize for barely register
Here is the number that should redirect a budget. Across those 379 B2B citations, g2.com and capterra.in were each cited exactly once. The review platforms that absorb the largest share of B2B software marketing effort are, as a direct source of ChatGPT citations, a rounding error.
This is not an argument that review sites are worthless. Buyers still read them, and they still matter for conversion once someone is comparing options. But if the goal is specifically to get named when ChatGPT builds its shortlist, another G2 campaign is close to the least effective thing you can do. The engine is not reading G2 to decide who to recommend. It is reading brand sites, tech press, and Reddit.
The 15% and 7% you are probably ignoring
Want to know how your brand scores on these same metrics?
We'll run 20 prompts across 3 AI platforms and send your report within 24 hours.
After brand sites, ChatGPT's next-largest source for B2B is independent publishers, led by TechRadar, with mainstream business press like Times of India and Economic Times behind it. Then community, which is almost entirely Reddit. Together, editorial and community are nearly a quarter of ChatGPT's B2B citations, and this is usually where a competitor who is beating you in AI answers is winning, on a channel you are not even contesting.
A "best CRM software 2026" roundup on a tech publisher, or a well-regarded Reddit thread comparing tools in your category, is a source ChatGPT reads and repeats. Most SaaS teams have a motion for review sites and no motion at all for earning tech-press inclusion or showing up credibly in the relevant subreddits. On ChatGPT's actual source diet, that is backwards.
The playbook, in priority order
Match the effort to the citation share:
- Own the brand-site layer (72%). Build extractable comparison pages, keep pricing and product facts machine-readable, and write for the "best [category] for [use case]" and "[category] alternatives" queries your buyers actually type. This is the biggest lever and the one nobody can take from you. We go deeper on the comparison-query motion in the CRM GEO playbook.
- Earn independent publisher coverage (15%). Get into the category roundups that real tech publishers write. This is slower than a review-site campaign and worth more to your ChatGPT visibility.
- Build real community presence (7%). Answer genuine questions in the subreddits your buyers use. Never hard-sell, because the community will bury you and the model will read the negative sentiment.
- Stop over-investing in G2 for AI visibility specifically. Keep it for buyer trust; do not expect it to move your ChatGPT number.
One honest caveat
Two caveats keep this honest. The numbers are the whole B2B software market pooled, not your product, and they show where ChatGPT looks, not proof that one comparison page vaults you into the answer. And this is ChatGPT alone; the five engines read very different source diets, so a plan built for it will not carry straight to Perplexity or Gemini. Before you move budget, run your category's prompts through each engine and see which of these sources is actually carrying the answer.
The one line to take to whoever owns the budget: ChatGPT builds its B2B shortlist from brand sites, tech press, and Reddit, in that order, and the review platforms most of that budget goes to are almost none of it.
Methodology. 379 ChatGPT citations across three B2B software categories (Conversational AI Platforms, CRM & Sales, HR & Payroll) in the August 2026 Cited Index, classified by source-domain type and pooled by citation count. Aggregate and market-wide: no individual product is named, and this is a different measurement system from the per-brand Citation Mix in the Cited dashboard. Shares are summed citation counts, not averages of per-brand ratios. This is Citation Source Mix, the sixth of the Cited 8 metrics.
Want to see how your own brand shows up across AI engines? Get a free AI Visibility Report across ChatGPT, Gemini, and Perplexity.