"SEO" for AI?

With AI being used frequently to find sites these days, was wondering how LLMs determine what sites to show in response to a query? Like if I ask "Give me good sites for cooking tutorials" how would it choose which results to show? thanks.

roidigh said:

With AI being used frequently to find sites these days, was wondering how LLMs determine what sites to show in response to a query? Like if I ask "Give me good sites for cooking tutorials" how would it choose which results to show? thanks.

Click to expand...

look, the second guy is just chasing headlines. that 60% reddit stat is a total hallucination— he probably saw a tweet about reddit's 40% citation frequency last year and turned it into a training data fact.

llms don't just copy reddit. they use reddit to gauge sentiment (do people actually like this site?) but for the actual facts, they’re scraping high-density data from the same big players we've always known.

if you want an ai to pick your site over the competition, you need to think about geo (generative engine optimization). it’s not about keyword stuffing anymore; it's about being machine-digestible.

here’s the real play:

1. kill the fluff. ai has zero patience for your 2,000-word backstory about your grandma’s kitchen. it wants fact density. if you can’t summarize the core answer in the first 80 words, the ai is going to skip you and cite the guy who did.
2. schema is your cheat sheet. use faq and recipe schema properly. it’s basically a json map that tells the ai exactly what to scrape so it doesn't have to work hard.
3. entity trust. stop obsessing over raw backlinks. start obsessing over brand mentions. if your site name shows up on reddit, quora, and niche blogs next to the words best cooking tutorials, the ai creates a semantic link. it doesn't even need a link to know you're the authority.

bottom line: the ai picks the site that's easiest to extract and most validated by real people. optimize for the bot, but get the human nodes (forums/blogs) talking about you.