Content Freshness: How Dates Affect Whether You Get Cited

Where freshness matters

Questions with a time component lean heavily on dates. "Best X in 2027," "current pricing for Y," "latest version of Z," and anything about a changing market tend to pull from recently dated sources. For these, a page dated two years ago competes poorly against a thinner page dated last month.

Questions about events, regulations, software, and prices behave the same way. The engine is trying not to give an outdated answer, and a recent date is its cheapest signal that the content reflects the present.

Engines with live web retrieval, like Perplexity and the browsing modes of ChatGPT, are more sensitive to dates than models answering from training data alone. The search layer underneath often has its own freshness logic, and the model inherits it.

Where freshness matters less

Durable questions do not reward recency much. "How does compound interest work," "what is a canonical tag," and "what causes seasonal allergies" have answers that do not change. For these, the engine cares about clarity and authority. A well-structured page from 2019 can keep getting cited for years.

Definitions, explanations of established processes, and reference material fall here. Refreshing them for the date alone is wasted effort.

The signals engines read

Several places carry a date, and they should agree.

• The visible date on the page, usually near the title.
• datePublished and dateModified in Article schema.
• The lastmod value in the XML sitemap.
• HTTP headers like Last-Modified, which some crawlers check.
• Dates mentioned in the text, such as "updated for 2027."

When these disagree, the engine has to guess. Schema says modified last week, the sitemap says last year, the page shows no date at all. In that case the model may fall back to the oldest or ignore the signal entirely. Pick one source of truth and generate the rest from it.

The date-bumping problem

The tempting shortcut is to change the date without changing the content. Some sites do this on a schedule. It tends not to work for long, and it carries a cost.

Engines compare the content they crawled last time with the content they see now. A page whose date moved but whose text did not is a page that claims to be fresh and is not. The signal gets discounted. If it happens across a whole site, trust in that site's dates drops generally.

The honest version is to update dateModified only when the content changes in a way a reader would notice, and to leave datePublished alone.

Making old pages read as current

For durable pages, there is a middle path. Keep the original publish date, add a visible "last reviewed" date, and actually review the page: check every fact, fix anything that drifted, and update examples that have aged. Set dateModified to the review date. The page now carries an honest recency signal without pretending to be new.

For time-sensitive pages, a real update is the only option. Rewrite the parts that changed, update the year in the title if it is there, and make sure the first paragraph reflects the present.

Showing dates at all

Some sites hide dates to avoid looking old. This usually hurts. A page with no date gives the engine nothing, and it may assume the worst. Show the date, keep it accurate, and let the content earn its citations.

How we handle it

At AEO GEO Labs, dates are part of the entity consistency check in every AI visibility audit. We look for pages where schema, sitemap, and visible dates disagree, and for pages that have been date-bumped without real changes. Then we separate the site's pages into durable and time-sensitive groups and set a review cadence for each, so freshness work goes where it has an effect.

A quick check

Pick three pages: one recent, one old and durable, one old and time-sensitive. Compare the visible date, the schema dates, and the sitemap lastmod. If they disagree on any page, that is the first fix. AEO GEO Labs offers a free schema markup generator that produces Article markup with correct date fields if you need a clean starting point.