Google’s John Mueller says AI training crawlers usually don’t give site owners a way to submit a sitemap, and he suggests a default file name or an RSS feed for sites that want those crawlers to find their content.
He made the comments with Martin Splitt in the Oct. 1 episode of Google’s Search Off the Record podcast, titled “Do sitemaps still matter?” He also said he has seen AI crawlers access his sitemap and RSS files in his own server logs.
Private Sitemaps And AI Crawlers
Mueller said site owners who want to keep a sitemap private can give it an unusual file name, leave it out of robots.txt, and submit it to Google directly. The downside is that other systems can’t find it, he said, and Bing would probably need its own submission.
AI training crawlers are different because “they usually don’t have any kind of a Console or any setup where you can submit a sitemap file.” For sites that want their content in AI systems, he suggested “either you stick to the generic naming, call it sitemap.xml, or you focus on RSS feeds.”
The sitemaps protocol says a robots.txt Sitemap line is independent of the user-agent line, so it isn’t tied to any one crawler’s rules. Feeds are easier to find, he noted, because they’re usually linked from a page’s HTML head.
“I’ve seen that happen in my server logs where some AI crawler accesses my sitemap file,” he said, and he’s seen the same with his RSS files. He didn’t name the crawlers and said he doesn’t know whether AI companies document this or what they do with the files.
Llms.txt Isn’t A Sitemap Substitute
Asked whether llms.txt could replace an XML sitemap, Mueller compared the Markdown file to an HTML sitemap and said Google’s systems can’t use it as a sitemap because it lacks the strict format.
“I think the hope is bigger than the reality,” he said, allowing that search systems might read Markdown files someday. “Currently none of this happens.” He’s fine with sites trying it but wouldn’t rely on it.
In August, he said the only crawlers on his test sites that claimed to accept Markdown were SEO tools. In May, I covered Google’s AI optimization guide listing llms.txt among the tactics sites don’t need for generative AI features.
Why A Valid Sitemap Can Show ‘Couldn’t Fetch’
Splitt closed by asking why Search Console reports “Couldn’t fetch” for a valid, public sitemap linked from robots.txt.
Mueller said one reason is host load. Google’s systems may be too busy to fetch the sitemap, and Search Console reports that as “Couldn’t fetch” too. The other is crawl demand, which can lead Google to skip the sitemap when its systems see no need to crawl much more from a site.
“And the crawl demand is very often based on the perceived quality of a website,” he said, adding that “it’s not purely a technical thing.” In February, he told a Reddit user that Google won’t use a sitemap if it isn’t convinced there’s new and important content to index.
Search Console’s Sitemaps report help page lists low crawl demand among the reasons a sitemap can’t be fetched and says better content brings more crawl demand. Other reasons it gives include robots.txt blocks, unresolved manual actions, and wrong URLs.
Why This Matters
A sitemap with an unusual name and no robots.txt entry stays hidden from systems it isn’t submitted to, according to Mueller, and AI training crawlers usually don’t offer a way to submit one. He suggested a default sitemap name or an RSS feed for sites that want AI crawlers to find their content. A robots.txt listing also tells crawlers that support the Sitemap directive where the file is.
For a valid sitemap showing “Couldn’t fetch,” both causes Mueller described, host load and crawl demand, sit outside the file itself.
Looking Ahead
Google’s Search Relations lead advised working from what’s documented now rather than planning around llms.txt. Server logs are where he saw AI crawlers reach his files, and they’re where you can check your own.
Featured Image: Tetiana Yurchenko/Shutterstock. Google wordmark: Source: Google.
