Neil Patel tests AI SEO tactics like llms.txt and content chunking and finds they don't meaningfully move traffic. He's skeptical of SEOs overselling these strategies and notes that measuring LLM-driven citations remains unclear.
By the way, SEOs like to invent work. Yeah, you know, I've been chunking my content in these llms.txt. That's how they get you to pay their consulting prices. Dude, so many SEOs say, oh, if you, if you don't do this thing, you're not going to get any LLM traffic. What's that? Cloudflare. Cloudflare ended up releasing something for LLMs, uh, where it ends up reducing the file size and makes it easier so they can just get to the text of the, or the meat of your content. The doc MD Maker? No. Is that what it was? What did you call it? The .md maker, like the markdown maker. There's a markdown— Markdown? Yeah. Because agents like reading markdown and to an extent, even HTML. Yeah. Markdown for agents. Yeah. Right. So they introduced this a while ago. Don't get me wrong. Like, it's cool, but whether you do it or not doesn't necessarily mean your traffic is going to go to zero or it's going to double up. And we've ran tests with this specifically using Cloudflare's solution with a few clients. And guess what the difference in LLM traffic was? Like almost nil? Pretty much, exactly. Now, I'm going to caveat this in which we don't have an exact way to know if you're getting more citations and people are seeing your brand mentioned more and it wasn't causing traffic. But from a traffic perspective, we didn't really see any meaningful results. And the reason I throw out that we're not 100% sure is you could say, oh, Neil, why aren't you using visibility tools? Well, with AI visibility tools, you put in a prompt and then it tracks it for you. Yeah. You don't know exactly what a user types in. There could be a sentence that Eric and I are searching for on Google's algorithm or some stats, but if we type it in with a few different words in that sentence, we can get different results. Heck, if you use the same prompt over and over again, you still can get different results. So the AI visibility tracking is not perfect.
