CiteSite AIAI search, GEO & SEO resources
Shorter pieces on what changes AI search visibility in practice, and on what the measurements do and do not support.
We gave our analytics chatbot less data on purpose
Raw database access made the chatbot more capable and less trustworthy. Why Ask CiteSite receives a small, computed brief instead, and what it refuses to answer.
Every Gemini citation we stored was the same wrong domain
Gemini reports its sources as links through Google’s own redirector. Read the URL and you record Google as the source — 768 times, then recommend outreach to an API endpoint.
Your page is fast for people and slow for crawlers
A browser waits for your page. An answer engine does not. The measurements that decide whether a crawler finishes reading you are not the ones on your performance dashboard.
A dashboard measuring the wrong rivals is worse than no dashboard
The competitive set is configured once and never looked at again, and every number is computed against it. How a tool reports 0% visibility against nine companies you have never met.
The canonical tag that points at a redirect
A site can serve every page correctly, declare a canonical on all of them, submit a clean sitemap, and still not get indexed. Here is the shape of that failure, found on our own site.
An AI mention is either a measurement or a memory
Ask the same question twice and you get two different answers. Only one searched the web. Telling them apart changes what an AI mention is evidence of.
A quotable paragraph is 25 to 110 words and never opens with “This”
A citation is one passage lifted onto someone else’s screen alone. The mechanical properties that decide whether yours survives the trip, or gets paraphrased instead.
AI search thinks your founder is your company
Most sites declare their own name three or four times, in markup that disagrees. What an entity actually is, and why the disagreement is the failure.
Which AI crawlers matter, and which ones you can safely block
There are three kinds of AI crawler and only one of them decides whether you get cited. Blocking the wrong line costs citations; blocking the right one costs nothing.
llms.txt: what it is, what it is not, and whether you need one
A plain reading of the llms.txt convention: what it does, what no engine has promised about it, and the case for publishing one anyway.
Schema markup for AI search: the types that actually get read
Which structured data an answer engine uses, which of it is decoration, and why the most common schema mistake is invisible in a browser.