Some listicles say Product Hunt, G2 and Crunchbase get you recommended by AI. Two studies say otherwise. What they found, and a test to run on your category.
Some 2026 listicles say ChatGPT, Claude and Perplexity check a short list of trusted hubs, such as Product Hunt, G2 and Crunchbase, so a listing on all three gets you recommended. Two studies bear on that claim. Neither supports it: in one, review sites were almost never the cited source, and in the other, every product was already listed, yet discovery questions surfaced them only 3.32% (ChatGPT) and 8.29% (Perplexity) of the time.
Where ChatGPT took its citations from
DerivateX's B2B SaaS citation study, published May 31, 2026, put one buyer question for each of 40 software categories into ChatGPT with web search, ten times each. That logged 233 recommendations across 219 distinct tools and 169 cited URLs.
ChatGPT attached a citation to 92.3% of the tools it named, so being cited was common. The question is where those citations pointed:
- Independent and niche blogs, plus vendor-owned content: 81.9%
- Major media and press: 8.8%
- Community and social sites, almost entirely Reddit: 8.4%
- Review aggregators (G2, Capterra, TrustRadius): 0.9%
G2 and Capterra each received zero citations across the 233 recommendations. The recommended tool's own website was the cited source only 11.6% of the time.
The study has clear limits. It covers one assistant, one kind of question (software buyers comparing options) and one point in time, and it does not look at Product Hunt or Crunchbase. DerivateX also says review profiles still shape buyer perception in other channels.
What happened to products that were already listed
The Discovery Gap, a paper submitted to arXiv on January 1, 2026 by Amit Prakash Sharma, randomly selected 112 startups from the top 500 products on the 2025 Product Hunt leaderboard. It ran 2,240 queries against ChatGPT (gpt-4o-mini) and Perplexity (sonar with web search).
Asked about a product by name, ChatGPT recognized it 99.4% of the time and Perplexity 94.3%. On discovery questions such as "What are the best AI tools launched this year?" the rates fell to 3.32% and 8.29%. The author reports that optimizing site content for AI visibility "showed no correlation with actual discovery rates". For Perplexity, referring domains did correlate with visibility (r = +0.319).
Every startup in this sample was already on Product Hunt, so the paper cannot compare listed products with unlisted ones. It shows that being listed did not lead to these products appearing often in discovery answers. It does not show that a listing is useless.
What the two studies add up to
Neither study found that a set of directories works as a gate into AI recommendations. In the first, review sites were almost absent from citations. In the second, every product was already listed, yet discovery questions surfaced them only 3.32% (ChatGPT) and 8.29% (Perplexity) of the time. Crunchbase is not covered by either, so we make no claim about it.
Taken together, they point toward being written about in current, specific comparison content, and, for Perplexity in the arXiv sample, building referring domains.
A worked example: test your own category
This is an illustration, not data. Say you sell scheduling software for clinics.
- Write ten questions a buyer might ask, such as which scheduling tools suit a small dental practice, or which ones work with a given booking system.
- Run each question in the assistants you care about, in a fresh chat, and save the answer.
- For every answer, note which tools are named and which pages are cited.
- Sort the cited pages into groups: your own site, independent comparison posts, media, Reddit and review or directory profiles.
Suppose the log shows your brand named in three answers out of ten, each time cited from the same independent comparison post, and no directory profile cited at all. That tells you the comparison post is worth more effort than another listing. If a directory profile does keep appearing as a cited source in your category, that is a reason to keep it current.
Before you run the test, check that the assistants can read your own pages. SerpCrest's AI Search Readiness Audit checks whether ChatGPT, Perplexity, Claude and AI Overviews can read and cite a site.
Where directory listings still fit
Our advice: keep the profiles a buyer would check when vetting a vendor, and judge them on that job, not on AI visibility. For the wider problem with bought listings, see why bundled directory packages can be full of filler listings. To compare directory types, see niche versus general directories.
