Layout testPlaceholder copy used to check typography, tables and spacing. Not a real article, and marked noindex.
Google updated the AI Overviews citation behaviour this month, and the practical consequence is smaller than the first-day takes suggested. The note now states plainly that an inaccurate lastmod value is ignored rather than penalised, which contradicts advice that has circulated for years. Here is what the note actually says, what we measured on three sites, and what is worth changing today.
What the note actually says
Two sentences carry the weight. The first confirms that lastmod is a hint, not an instruction. The second says that if the value is demonstrably wrong — a date that moves every time the page is served, for instance — the crawler stops trusting the field for the whole sitemap. That second half is the part worth acting on: the risk is not a penalty, it is being ignored.
Nothing in the note suggests a ranking effect, and nobody at Google claimed one. The mechanism is scheduling: a trustworthy signal earns a slightly faster revisit, an untrustworthy one earns nothing at all.
What we measured
We pulled fourteen days of server logs from three sites before and after correcting their lastmod handling, and counted the median time between a page changing and Googlebot fetching it again.
| Site | URLs in sitemap | Recrawl before | Recrawl after | Change |
|---|---|---|---|---|
| A — daily news | 1 240 | 4.1 days | 3.8 days | −7% |
| B — documentation | 860 | 11.2 days | 10.9 days | −3% |
| C — ecommerce | 15 400 | 6.7 days | 6.6 days | −1% |
The effect is real, consistent in direction, and small. On the news site it saves roughly seven hours; on the shop it is inside the noise. Anyone promising a step change from a sitemap edit is selling something.
Where it does matter
- Sites publishing several times a day, where hours of delay are visible in traffic.
- Large sites where crawl budget is genuinely constrained and wasted fetches cost real capacity.
- Any site currently sending a
lastmodthat updates on every request — that one is worth fixing this afternoon.
A signal you cannot support is worse than no signal, because it costs you the field for everything else in the file.
FAQ
Does a wrong lastmod get a site penalised?
No. The documentation says the value is ignored, and the field can stop being trusted for the whole sitemap. That is a lost opportunity, not a penalty.
Should lastmod change when only a template changes?
No. Reserve it for changes to the page’s own content. Site-wide template edits touching every date is exactly the pattern that makes the field untrustworthy.
Is IndexNow a better option?
For Bing and Yandex, yes — it is a push, not a hint, and it is nearly instant. Google does not consume IndexNow, so both are worth having.
Correction, August 27, 2026: an earlier version of this test article said the recrawl improvement on site C was 4%. Recomputed from the same logs, it is 1%. The conclusion is unchanged.
