After 1,339 submissions (2026-08-21 to 2026-10-12) the count is zero. Not one new tool was named in 6,710 answers.

The obvious objection: gpt-4o-mini answers from training data, and most of these tools launched after its cutoff. So we took a seeded random sample of 125 of them and re-asked the same questions with live web search switched on.

Still zero. 0 of 125, across 625 answers and 2,079 source citations.

What the search-enabled model did instead

It named about 12 products per answer (vs ~6 without search), so it had more room. The extra slots went to incumbents too: ChatGPT (54% of sampled tools' answers), Gemini (44%), Canva (43%), Claude (38%), Adobe Firefly (31%).

The useful part for anyone launching something is what it read vs what it cited:

• Reddit is read, not cited. A Reddit thread was retrieved for 95% of answers and cited in 8 of 625.

• Directories barely register. Futurepedia, There's An AI For That, Toolify, FutureTools and AlternativeTo were cited in none of 625 answers. Our own directory was retrieved 9 times and cited zero times. (Yes, we're publishing the number that makes us look bad.)

• What does get cited: the incumbents' own docs (openai.com, help.openai.com, learn.microsoft.com, helpx.adobe.com) and a handful of review publishers (technologyadvice.com, techradar.com, g2.com).

• Search reached one sampled tool's own site twice, read it, and still didn't name it.

101 of the 108 sampled tools with a readable domain date were registered after the model's training cutoff, so these are exactly the tools live search is supposed to surface.

What I'd take from it if I were launching

1. A directory listing is fine for people browsing that directory. In this data it is not a path into a ChatGPT answer.

2. The cited sources are review/comparison publishers and documentation-shaped pages. Getting written up on one of those likely matters more than another listing.

3. "0 of 125" isn't "never". The exact 95% upper bound is ~3% of new tools per five questions. It's rare, not impossible, which means it's worth measuring for your tool rather than assuming.

Full report, methodology and per-category tables: ChatGPT named 0 of 1,339 new AI tools

If you've shipped an AI tool and want to see the five buyer questions for your category and who gets named instead of you, the scan is free (no signup): run the free AI visibility scan. There's a paid weekly re-check if you want to track it over time; the scan alone answers the question.

Happy to answer methodology questions in the comments. The matcher, aliasing and the two false positives we threw out are described in the report.