There is a reassuring thought in the SEO world: get your technical side in order and visibility follows. Neat titles, a fast site, structured data, and the rest takes care of itself. We have a measurement that roughly knocks that thought down.
The painful figures
A site from our own measurements scores 90 out of 100 on technical AI readiness: crawlers welcome, titles in order, everything readable. That site appeared in zero of 225 AI answers in its category. Not a single assistant, not a single question, not a single draw.
In the same category there is a platform that flatly refuses our crawler with a 403 error. Technical score: substandard. That platform appeared in half of the Claude answers and dominated the citations.
Why this holds (however bitter)
AI assistants do not build their answer on your site. They build it on the sources they trust for that question: comparison sites, lists, trade media, directories. Your site is at most one of the candidate sources, and for a choice question ("which agency do I choose?") an assistant rather cites a comparison than an interested party.
So technique is a precondition, not a strategy. A site that blocks crawlers cannot be read, and you have to fix that. But beyond that the effect stops: going from 80 to 95 on technique changes nothing about the answer of an assistant that does not use your site as a source anyway.
Where visibility is actually decided
In our measurements the list of cited sources per category is surprisingly stable: a handful of domains feeds the majority of the answers. The question that matters is not "is my site perfect?" but "am I on the pages the assistants cite from?" That is a list of names and URLs, not an abstract score.
That is why Geony ranks source actions above technical actions: first the mentions on the sources that determine the answer, then the rest. The reverse order feels more productive, but our data says it is not.