Freshness signals: what dateModified really does for SEO
Changing a date stamp doesn't fool Google, and AI engines weigh recency differently again. Here is what freshness signals actually do and how often to update.
Freshness is a real ranking factor, but it only applies to certain queries, and most sites misuse it. Editing a page's date without changing anything else rarely helps, and it can quietly cost you trust. Here is what actually happens when you update a date, and how AI engines weigh recency differently than Google does.
When freshness actually matters
Google has used some version of query deserves freshness (QDF) since 2007. It kicks in for queries tied to a moving target: current events, recurring events, product releases, prices, or anything people search with words like "today" or "2026" baked in. For a fixed reference topic, like how a protocol works or what a term means, freshness barely factors in. Depth, accuracy, and links do more work there.
- Freshness matters more: news, pricing pages, software version guides, best-of and roundup lists, anything tied to a season or event
- Freshness matters less: definitions, how a protocol or standard works, historical explainers, evergreen reference material
datePublished vs dateModified
On Article and BlogPosting schema, datePublished and dateModified are separate fields for a reason. datePublished should never move. dateModified should only move when the substance of the page changes: new facts, a corrected error, an added section, updated numbers. Google has said its systems try to detect date stamps that update without matching content changes, and pages caught doing this can lose trust rather than gain a ranking boost.
A new date on old content is a promise. Search engines and readers both check whether you kept it.
The visible last-updated date
A human-readable "last updated" line on the page matters as much as the schema field, because it's what readers and AI systems quoting your page actually see. Keep it in sync with dateModified. A page that displays "updated September 2026" while the structured data still says March creates a mismatch that undermines both.
- Match the visible date to the dateModified value exactly
- Keep the original publish date visible too, since it signals how long you've covered the topic
- For a significant update, say what changed in a line or two, not just a new date
- Don't restart the publish date on a URL that already has history and links pointing to it
How AI engines weigh recency
Google's freshness scoring has been tuned for close to two decades. The retrieval systems behind ChatGPT, Perplexity, and AI Overviews are younger, and none of them publish how heavily they weight recency. What practitioners have observed, without hard data to back it up, is that these systems seem to favor pages that look recently verified when a query itself signals urgency, such as "current" or a specific year. Treat this as an early, unproven pattern worth watching, not a confirmed ranking rule. Don't restructure a whole content strategy around it.
How often to update, by content type
- Pricing, news, and anything tied to a live fact: update the moment the fact changes, not on a schedule
- Tutorials and how-to guides: check every quarter, or immediately after the tool, API, or interface they describe changes
- Roundups and best-of lists: refresh whenever an entry is added, removed, or clearly out of date
- Evergreen reference and definitional content: review once a year for accuracy, even if nothing needs to change
The simplest audit: pull every page with a visible last-updated date older than a year and ask whether the content underneath actually still holds up. If it does, leave the date alone. If it doesn't, fix the content first and let the date follow, not the other way around.