Field note
Recommended ≠ cited: two different games
From one audit I built two lists: who the engines recommend, and who they cite. They barely overlapped.
From a single audit I pulled two different lists. One: the companies the AI assistants recommended — named as the answer when a buyer asked who to use. The other: the sources they cited — the URLs underneath those answers. I expected heavy overlap. There was almost none.
The provider recommended most often across the questions barely appeared in the cited-source list at all. And the source cited most — by a wide margin, Reddit — isn't a "provider" anyone gets recommended as. A company can be the thing every answer is built from and still never be the thing recommended; another can be recommended constantly while its own pages are nowhere in the citations, because third parties carry it.
That's because they're two different games. Citation is machinery — being retrievable and quoted inside the answer (the GEO half). Recommendation is the outcome — being named as the pick (the fourth dimension). One feeds the other, but they don't move together, and a fix for one isn't a fix for the other.
The practical consequence: don't measure only "are we cited?" or only "are we recommended?" — measure both, because they're different signals pointing at different work. Treating them as one number hides which problem you actually have.
The honest caveats: this is one audit, one category, one day — the split between the two lists is the finding, not the exact names, which shift on a re-run. But the split itself is the point: being a source and being the answer are not the same achievement.
This is the kind of distinction an audit is built to separate — see also do engines cite the same sources?
New here? Start with the overview → · By Denes Csaszar · written in the open