Google says that a site of around 500 pages or fewer, well linked internally, can do without a sitemap, and John Mueller still recommends leaving it enabled: it costs nothing and most CMSs produce it on their own. Only the address and the modification date count; priority and changefreq are ignored, and the date is only read if it is accurate. A “Couldn't fetch” status on a valid file often comes down to crawl demand, and therefore to the site's perceived quality. As for llms.txt, it is not submitted like a sitemap, and according to Mueller no AI system uses it.
Contents
On October 1, 2026, Martin Splitt hosted John Mueller for an episode of Search Off the Record devoted to sitemaps, titled Do sitemaps still matter?1 The podcast belongs to Google Search Central, Google's official site for site owners and SEO professionals.2 Splitt is a developer advocate there and Mueller leads the Search Relations team.3 The starting point is the mail Splitt receives: site owners worry about what they see in Search Console about their sitemap, and the episode reviews what Google expects from a sitemap in 2026.
01Does a small site need a sitemap?
It is the question every small site owner ends up asking.
What the documentation says
The documentation lists the sitemap among the aids to discovering pages. It mainly serves large sites, where linking every page becomes hard, new sites that receive few links, and sites that publish videos, images or news.4
A site of 500 pages or fewer, whose important pages are all reachable from the home page, may not need one, it specifies.4 That reachability condition depends on how internal links distribute the important pages from the home page.
What Mueller says
Mueller agrees: small sites probably do not need a sitemap, since Google can crawl them without help. He does point to a real problem, though: the threshold. Ten pages is small. Fifty, people hesitate.1
His usual advice is therefore simple: enable it. Almost every CMS, including static site hosts, generates a sitemap by default unless you turn it off. If the site grows, the file is already in place, and if it stays small, it bothers nobody.1
When it becomes really useful
Its value grows with the volume of content and the pace of changes. Mueller takes the example of a store: without a sitemap, Google has to find through regular crawling the product page whose price just changed, and sometimes gets there too late. With a sitemap, the modification date points it directly to the page to revisit.1
It is one of the first settings to check in the SEO of an online store.
News sites are the clearest case. Google provides a dedicated sitemap for them, which only lists articles from the last two days and does not exceed 1,000 news tags.5
02The address and the date: everything else has gone
The format has slimmed down a lot since its early days. Two useful pieces of information remain per page, and Google expects only one thing of them: that they be accurate.
Two abandoned fields
In the episode, Splitt and Mueller explain why priority and changefreq disappeared. Ask SEOs for the priority of their pages: everything is at the maximum. Ask for the change frequency: everything changes all the time, especially since a server-rendered page is technically recalculated on every visit.1
As a result, these two fields taught Google nothing. The documentation says Google ignores them.6
The date: only read if it is credible
Mueller describes the date as that of the page's last significant modification. If you show today's date on all your URLs, Google notices and stops trusting it.1
No spam system classifies the file as bad for all that: Google falls back on the URLs themselves and processes each new page that appears.1
The documentation says the same thing, more precisely. The value must reflect the last significant update, date and time, and Google only uses the field if the dates are consistently accurate. Changing the main content, structured data or links counts. Changing the copyright year does not.6
Mueller warns that this is his current understanding, that it may have changed, and that Google has an ambivalent relationship with this field, because many sites fill it in poorly.1
In practice, the date must come from the actual modification of the content and never from the time the file was generated. A sitemap regenerated every night with today's date falls back into the case described above.
List the canonical addresses
The sitemap is for declaring addresses as you want them indexed. Mueller takes the example of a URL to which a date is appended to force the latest version to display: Google will try to index the version without the date as canonical.1
He adds, still according to his understanding, that listing a URL in the sitemap makes it slightly more likely to be chosen as canonical among several neighboring addresses.1 The documentation asks for absolute URLs, with the protocol, and a UTF-8 encoded file.6
| Field | Read by Google? | Condition |
|---|---|---|
| Page URL | yes | absolute address, canonical version |
lastmod | yes, if reliable | last significant modification |
priority | no | ignored |
changefreq | no | ignored |
| Language versions | yes | can be declared in the sitemap |
| Images and videos | supported | less decisive than before |
Mueller mentions in passing that language versions can also be declared in the sitemap, which the documentation lists among the accepted information.4
For images and videos, dedicated sitemaps are still supported, but he doubts how much they matter today: Google recognizes this content directly on pages better than before.1
03Formats, limits and file location
The accepted formats, the size limits and where to declare the file: the essentials fit in a few points.
Accepted formats
Google reads XML sitemaps, RSS and Atom feeds and text files. XML is the most versatile, because it can also carry image, video and news data.6
Mueller considers an RSS feed very close to a sitemap. It lists recently modified URLs with a date, often capped at ten or twenty entries, and it is submitted in Search Console like a sitemap.1
Its advantage shows on a large site: instead of going through hundreds of sitemap files one by one, a system reads the feed and sees at a glance the pages that have changed.1
Size limits
A sitemap file does not exceed 50,000 URLs or 50 MB uncompressed.6 Mueller quotes these figures from memory and specifies that the size limit applies to the decompressed file: you can compress with gzip, but you cannot aim for a compressed size.1
Beyond that, URLs are spread across several files tied together by an index file, which can list up to 50,000 sitemaps. Search Console accepts up to 500 index files per site.7 The two hosts conclude that a site that gets to several index files has enough URLs to justify a sitemap anyway.
Where to declare it
The file can have any name, but sitemap.xml remains the one many systems guess on their own. To make it known, you declare it in robots.txt or submit it in Search Console.6
A confidential name submitted only in Search Console keeps the file discreet, but Bing and the other engines then have to be notified separately.1
AI crawlers generally have no console to submit a file to. Mueller has seen in his server logs that some read his sitemap and RSS feeds, without this behavior being documented and without his knowing what they do with them.1
An HTML site map replaces nothing
An HTML site map is a map of the site intended for visitors. Google follows its links like any others, which helps crawling, but it cannot treat it as a sitemap, for lack of a strict structure, and it cannot be submitted.1
Mueller adds that an HTML map does not list everything: on a store, it shows the categories, rarely every product.1
04“Couldn't fetch” even though the file is valid
This is the episode's starting point. The file validates, it is public, it is declared in robots.txt, and yet Search Console shows “Couldn't fetch”.8
Mueller says he sees this case come up often in the help forums. He gives two main causes, both linked to what Google allows itself to crawl on a site.1
Google did not have time
There is a limit to what Google crawls on a host. If its crawlers are busy with other pages, the sitemap is not fetched, and Search Console shows it as couldn't fetch, for lack of time to do so.1
Google sees no point in it
Second cause: crawl demand is low, and Google skips the file because it already has what it needs. This demand very often rests on the perceived quality of the site.1
If the systems judge the overall quality as average, they spend few resources crawling and indexing, and so do not bother with the sitemap for now. If quality improves markedly over time, they will end up using it.1
The Search Console documentation confirms this reading. Among the causes listed for this status is low crawl demand, with this clarification: the higher the quality of the site's content, the stronger the crawl demand.8
Ordinary causes to rule out first
Before concluding there is a quality problem, you need to rule out the common causes, which the same documentation page lists: a sitemap blocked by robots.txt, an unresolved manual action on the site, an incorrect sitemap address returning a 404 error, occasional server errors.8
It recommends opening the sitemap's details in the report, then testing the address with the URL Inspection tool.8 These crawling and indexing checks also make up the first block of a do-it-yourself technical audit with a checklist.
For a site of a few dozen well-linked pages, remember that the documentation exempts it from needing a sitemap, and that Google keeps discovering pages through links.4
05llms.txt does not replace the sitemap
Splitt asks the question head-on: will the XML sitemap disappear in favor of llms.txt? Mueller starts by describing the file. It is a markdown document that links to content, a bit like an HTML site map, in the hope that an AI system will use it to understand the site better.1
In his view, the hope exceeds the reality, even if that may change. Google's systems cannot use it like a sitemap, for lack of a strict format. It might one day be read as markdown whose links get followed, but for now nothing of the kind happens.1
His advice: focus energy on what works and what is documented, rather than on a hoped-for future state. If your CMS produces an llms.txt on its own, fine, but you should neither rely on it nor make it a priority.1
His public statements point the same way. In June 2025, he wrote on Bluesky that no AI system was using llms.txt, and invited people to check their server logs: chatbots fetch pages there, never this file.9 In June 2026, on Reddit, he called the file purely speculative for now.10
As for the XML sitemap, Mueller does not see it disappearing: its exhaustive nature keeps it useful.1