Measured

Where AI answers about software actually come from

Everyone in this field will tell you whether an assistant names you. Almost nobody will tell you which pages the answer was built from, which is the half you can act on. Here is ours, including the part where we lose.

What we did

We asked the same 21 buying questions every day from 20 July 2026 to 26 July 2026, in our own category. Each question got one live web search, and the identical results went to ChatGPT, Gemini and Perplexity, so the models differ in how they read the web and never in which web they saw. 19 of the 21 were answered by all three; on the other 2 Gemini returned nothing, so those are two models.

Sources are counted per answer, not per mention. A page one answer leaned on three times is one answer’s worth of evidence. Counting the repeats would measure one model’s verbosity instead of the category’s citation surface.

The 93 sources, ranked

How many of the 21 answers each domain was a source for. The full set was 93 distinct domains, so this is the head of a long tail.

• 12reddit.com
• 7rankability.com
• 6frase.io
• 6semrush.com
• 5seranking.com
• 5tryprofound.com
• 4alhena.ai
• 4dageno.ai
• 4gracker.ai
• 4nightwatch.io
• 4useomnia.com
• 4wpengine.com
• 4youtube.com
• 4zapier.com

Reddit, by a distance

reddit.com was a source for 12 of 21 answers, 57%, more than twice the next domain. That is not a surprise once you look at what a buying question needs: people comparing named products in their own words, including the failure modes a vendor page will never print. See does Reddit affect AI search for the same finding across three other categories.

The practical catch, which most advice skips: the subreddits in a software category mostly ban self-promotion outright. We read the published rules of ten of ours and the majority forbid posting your own product at all, several are private, and the permissive ones allow it only when the post carries value on its own. So Reddit is the largest source and the one you cannot buy your way into, which is precisely why it holds.

Below Reddit, it is other vendors’ blogs

Look at the rest of the table. Almost all of it is software companies running “the best tools in this category” posts, and one is a direct competitor’s own site. Those posts are being read back to buyers as answers, including to the prospects of the rivals listed inside them.

That makes them the cheapest target on the page. You do not have to outrank them. You have to be in them, which is a different and much shorter piece of work than a content programme.

Where we are in our own table

Nowhere. Our domain is not in the 93, and across the 21 answers we were named zero times. We publish that because a vendor showing a table that flatters them is an advertisement, and a vendor showing a table they lose in is evidence the table is real. It is also the honest state of a young product in a category that has been written about by larger companies for longer.

It is the reason we think the source list matters more than a visibility score. “You are not named” gives a reader nothing to do. “57% of the answers in your category are built from Reddit, and the next six are vendor roundups you could be listed in” is a decision about where the week goes.

Questions