Which Publications Do AI Engines Cite Most?
Citation studies show a small set of domains dominating AI answers. Here is what the data says and what it means for press strategy.
If AI answers drew evenly from the open web, press strategy would not change at all. They do not, and that is the finding worth acting on.
Citations are concentrated
Across tracking studies, roughly 30 domains capture about 67 percent of citations within a given topic. Two thirds of the sources, from a few dozen sites.
That concentration is the opposite of how the open web works, and it means the question is not whether you have coverage. It is whether your coverage sits on one of the domains that engines actually return to.
Who is on the list
Historically Wikipedia and Reddit, each around 12 to 13 percent of ChatGPT citations in the US. After the shift in late 2025, professional publishers and professional networks took a much larger share. Forbes doubled. LinkedIn gained. PR Newswire posted the biggest increase of any news domain.
The pattern across studies is consistent: established publishers, reference sites and professional networks, rather than blogs, forums or company websites.
What that changes about press
It makes the choice of publication more consequential than the existence of coverage.
A feature on a small site with no history is still a credential for a human reader. It is close to invisible to an engine that keeps returning to the same few dozen domains. A feature on a masthead inside that set does two jobs at once.
This is also why tier 1 and tier 2 publications are worth distinguishing properly rather than treating all coverage as equivalent. The gap was always real for human readers. It is sharper for machines.
What not to conclude
That buying one article on a big domain makes you appear in AI answers. It does not work that way.
Engines weigh corroboration, recency and relevance alongside domain authority. A single piece is one signal. Several consistent pieces across recognised publications is a pattern, and a pattern is what gets reproduced in an answer.
The honest caveat
These numbers move. The Reddit collapse took about a month, and it was driven by commercial agreements rather than anything a marketer did. Treat any citation ranking as a snapshot rather than a law.
The durable part is the concentration itself. Whichever domains sit at the top in a year, there will be a small number of them.
Getting started
Publisive places clients in the tier 1 and tier 2 publications these studies keep surfacing. We write the piece to your brief, you approve every word, and it stays findable.
We write and place features in the publications our clients choose. About us
Frequently asked questions
01Which domains do AI engines cite most?
Studies consistently find reference sites, professional networks and established publishers. Wikipedia has historically been the largest single source, and after the 2025 shift toward professionally produced content, publishers such as Forbes gained substantially.
02Does volume of coverage matter more than placement?
Placement matters more. Twenty articles on sites with no search history is a weaker position than three on domains inside the cited set, because citations concentrate in a few dozen places rather than spreading across the web.
03Will these rankings stay the same?
No. One licensing agreement reshaped ChatGPT's source mix in about a month. Treat any citation ranking as a snapshot. The durable finding is the concentration itself, not the specific names.
