Guides
How answer engines pick the sources they quote, which AI crawlers cost you something when you block them, and what the published research actually supports — with the limits of each claim stated.
How answer engines choose which sources to quote
Answer engines assemble responses from passages they can lift and trust. Here is what that means in practice, what the research found, and what nobody can honestly promise you.
Read guide 8 min readTraining vs. retrieval AI crawlers: which ones you can safely block
GPTBot governs OpenAI training crawls; OAI-SearchBot governs ChatGPT search discovery. They are different crawlers, the names are similar, and a rule written for one routinely catches the other.
Read guide 7 min readWhat is llms.txt, and is it worth publishing?
A markdown file at your site root that tells a language model what your site covers. The format is simple and settled. Whether any engine reads it is not — and most write-ups skip that part.
Read guide 10 min readWhat a liftable answer block is, and how to write one
The opening paragraph that answers the question on its own: the exact rule Crawlune's audit applies, how much it weighs, and four rewrites that show the difference.
Read guide 11 min readWhat the GEO research found, and what it did not test
The one peer-reviewed study most GEO advice rests on, read closely: how the experiment worked, what its headline number measures, what it never tested, and what survives the caveats.
Read guide