lastmod: definition and why it proves little

lastmod is a tag inside each <url> entry of an XML sitemap. It holds a date, and sometimes a time, that the site owner says marks the last change to that URL.

<url>
  <loc>https://example.com/pricing</loc>
  <lastmod>2026-10-14</lastmod>
</url>

The key word is declared. The crawler reading the sitemap has no way to confirm the date from the tag itself. It is a claim, in the same category as a visible "Updated on" line or a schema dateModified field.

Most sites never type a lastmod by hand. A CMS or sitemap plugin writes it, usually from the database timestamp of the last save.

That creates three common patterns:

• Accurate: the date moves only when the content changes.
• Noisy: the date moves on every save, template change or plugin update, so every URL looks fresh.
• Stale or missing: the plugin never updates it, or omits it.

You can't tell which pattern a site follows from the tag. You can only tell by comparing it with other date signals on the same page.

It matters less than most GEO advice claims. Anthropic's web search tool exposes a page_age field for each result, described as when the site was last updated. Exposure is not weighting. No vendor states that the date affects which pages get selected.

The best-known supporting study is observational. A Seer Interactive study (July 2026) found 75% of cited pages had been updated within a year. It had no control group of uncited pages and read dates from publisher-declared fields, including lastmod. If most of the web carries a recent CMS-generated date, that figure tells you nothing.

One lab result points the other way. The 252,000-trial factorial study found a recent timestamp helped, but only for documents already placed in the model's context. That says nothing about whether a page gets retrieved in the first place. Treat freshness as unresolved.

What lastmod can do is expose contradictions. If the sitemap says October, the page says March and the header says last year, something in your publishing pipeline is broken. The date consistency check playbook walks through that comparison step by step.

"Update lastmod and AI will cite you"

No evidence supports this. Changing a self-declared date without changing the content is indistinguishable from spoofing, and it corrupts your own records. The question page on changing the last updated date covers this in more depth.

"A fresh lastmod fixes outdated AI answers"

If ChatGPT or Claude describes old facts with search off, the cause is the model's knowledge cutoff. No sitemap date reaches a retrieval-off answer. For answers that retrieve live, the fix is publishing the correct fact on a page crawlers can fetch. See what to do when AI describes dropped features.

"Pinging IndexNow with lastmod speeds up AI citation"

The IndexNow specification never mentions AI, and a 200 response only confirms receipt. The comparison of IndexNow and XML sitemaps explains what each one actually does.

• dateModified: the schema.org property that declares a page's last change. It is also self-declared.
• Last-Modified header: the HTTP response header the server sends. It often reflects file or cache changes, not content edits.
• Visible date: the "Updated on" line a reader sees. It is the one date a human can check against the content.
• page_age: the field Anthropic's web search tool returns for each result.

When all four agree and move only on real edits, your dates are trustworthy. When they disagree, fix the pipeline before reading anything into them.