Content freshness: why AI search leans toward what's current (and crawled)

AI answers cite what's been crawled, and they lean toward pages that are kept current, so a refresh that search engines haven't re-crawled does nothing for you yet. This guide covers what freshness means, when it matters, what counts as a real update, how to signal changes honestly, and how to make sure engines see them.

What content freshness means to search engines

To a search engine, content freshness means the stored copy of a page reflects current, accurate information and changed in substance when the topic changed. Engines learn about changes by recrawling. Until then they serve, rank and quote whatever they last stored.

Freshness signals come in three kinds:

• Discovery signals. Sitemap lastmod, internal links from frequently crawled pages and, on Bing, IndexNow pings tell a crawler a URL is worth revisiting.
• On-page signals. Updated facts, figures, screenshots, a visible "last updated" date and dateModified in structured data.
• Comparison. When an engine recrawls a page, it compares the new version with the stored one. That's why a new date on unchanged text doesn't fool anyone for long.

When recency matters for a query, and when it doesn't

Recency matters for queries where people expect current information, such as news, launches, prices, software versions and "best X" lists, and matters far less for evergreen questions. Google says its "query deserves freshness" systems are "designed to show fresher content for queries where it would be expected." Its examples are reviews of a newly released movie and news after an earthquake (Google ranking systems guide).

So does Google favor fresh content? Only when the query calls for it. A definition of "crawl budget" doesn't go stale every month. A "best project management software" roundup or a pricing comparison does.

Research on AI answers points the same way. A July 2026 audit of four Chinese generative engines, covering 160,860 citation records, found cited pages aged faster for time-sensitive queries (a half-life of about 39 days) than for less time-sensitive ones (about 68 days) (arXiv 2607.15771).

Why AI answers lean toward current sources

AI answers lean toward current sources because they're built from search indexes that already rank fresh pages higher for time-sensitive queries, and possibly because the language models themselves read dates as relevance. The lean is real but modest, and it varies by engine.

The main vendor studies:

• Seer Interactive (July 2026). The study covered 7,683 pages and 47,097 citations on ChatGPT, Gemini and Perplexity for four brands, from March to June 2026. 75% of cited pages had been updated within a year. On the pages where both dates were readable, 72% had been updated within a year but only 42% had been published within a year (Seer Interactive).
• Ahrefs (July 2025, about 17 million citations). AI assistants cited pages that were younger than Google's organic results on average. Even so, the average cited page was about 2.9 years old, and Google AI Overviews cited slightly older pages than organic results (Ahrefs).
• SE Ranking (November 2025, 216,524 pages). Pages updated in the past three months averaged 6.0 ChatGPT citations versus 3.6 for outdated pages. Brand-new pages averaged 3.6 too, so newness alone showed no edge (SE Ranking).

On the model side, a lab study gave passages fake publication dates and had seven LLMs rerank them. Every model pushed newer-dated passages up, and up to about a quarter of choices between equally relevant passages flipped toward the newer date (Fang et al.). The dates in that test spanned a century, and it hasn't been repeated on production engines. For how language models pick and describe sources more broadly, see our LLM SEO guide.

Keep it in proportion. Authority and relevance predict citations better than freshness does, and Google says AI features run on its core ranking systems. Freshness looks like a tie-breaker that helps a maintained page stay in the running for ranking in AI Overviews and for getting cited in ChatGPT.

The invisible-update problem: you updated it, Google hasn't re-crawled it

The invisible-update problem is the gap between when you publish a change and when search engines fetch it. During that gap, Google Search, AI Overviews and AI Mode work from the old copy. Google says crawling "can take anywhere from a few days to a few weeks" and that requesting a recrawl repeatedly "won't get it crawled any faster" (Google Search Central).

Google is explicit that its AI features draw on "relevant, up-to-date web pages from our Search index" (Google AI optimization guide). "Up to date" means up to date in the index, not on your server. That's how a pricing page you fixed in March can still be quoted at the old price in April.

The gap hurts most where freshness matters most:

• Pricing and plan changes.
• Product launches and feature releases.
• Comparison pages after a competitor changes.
• Statistics pages when a new year's data arrives.
• Refreshed guides you expect to win back lost traffic.

What counts as a meaningful update (and what doesn't)

A meaningful update changes what a reader learns from the page: new facts, corrected figures, new sections, current screenshots, or removed outdated advice. Changing the date, swapping a few words or adding "2026" to the title doesn't count.

Google asks site owners directly, in its helpful-content self-assessment: "Are you changing the date of pages to make them seem fresh when the content has not substantially changed?" (Google Search Central). Its guidance for news publishers is just as plain: "don't artificially freshen a story without adding significant information or some other compelling reason for the freshening" (Google Search Central Blog).

Fake freshness is common. In a July 2026 vendor audit of 370 page-one Google results across 50 keywords, On-Page.ai compared archived versions and found that 46% of the pages claiming an update in the past year had changed nothing but the date (under 2% of sentences different). The audit also found no statistically significant freshness difference between pages AI Overviews cited and pages they didn't (On-Page.ai). Moving the date isn't a proven win, and it risks Google trusting your dates less.

dateModified, sitemaps and lastmod, done honestly

dateModified and sitemap lastmod should record when a page's content last changed in a meaningful way, and nothing else. Used honestly, they help crawlers spend their visits on pages that changed. Used to fake freshness, they teach engines to ignore your dates.

Bing is specific: "Avoid setting lastmod to the time your sitemap was generated unless the content on that URL was actually updated." The value "should reflect the true last modification time of the page content" (Bing Webmaster Blog).

How to set dates up honestly:

1. Store a content-changed timestamp in your CMS that moves only when an editor marks a substantive edit, not on every save or deploy.
2. Use that timestamp everywhere: sitemap lastmod, dateModified in Article JSON-LD, and the visible "last updated" line. Google asks that structured data match the visible page.
3. Keep datePublished fixed. Republishing an old URL as "new" confuses readers and engines.
4. Add a short "what changed" note on major refreshes. It tells readers and crawlers why the date moved.
5. Audit for deploy-time lastmod. Many sitemap generators stamp every URL with the build time. If every page in your sitemap changed today, none of the dates mean anything.

A refresh cadence that beats content decay

Content decay is the slow loss of traffic, rankings and citations as a page's facts age and competitors publish newer answers. The fix is a refresh schedule built around your most important pages, not a higher publishing volume. Seer's 2026 study points the same way: the pages that held citations were established pages kept reasonably current, not a stream of new posts. Pages cited in all four months of its study were actually less recently updated than pages cited in just one month (68% updated within a year, against 86%) (Seer Interactive). No study has shown that publishing more often earns more AI citations.

A content refresh routine that works in practice:

• Monthly: pages with prices, plans, release notes or "best of" lists, plus anything about fast-moving platforms such as AI search.
• Quarterly: your top 20–50 pages by revenue, traffic or citations, and any page whose Search Console clicks fell sharply quarter over quarter.
• Yearly: evergreen guides and definitions. Check facts, links and screenshots, and change the date only if the content changed.
• On trigger: a competitor relaunch, a product release, a new data source, or an AI answer that quotes you incorrectly.

For product teams, the trigger list is mostly your own releases. See freshness for SaaS releases. These cadences are working practice, not research findings. Adjust them to how fast your topic actually moves.

Get updates re-crawled and check what Google saw

A refresh is finished only when search engines have fetched the new version, so close every update with a re-crawl request and a check. Treat "published" and "seen" as two separate statuses.

1. Publish the substantive change and update the content-changed date.
2. Make sure the URL is in your sitemap with the true lastmod, and linked from a page Google crawls often.
3. Ask Google to recrawl. Use URL Inspection for a handful of pages, a sitemap for many. Our guide shows how to get Google to recrawl a page.
4. Notify Bing. Bing recommends IndexNow for real-time URL submission (Bing Webmaster Blog). See our Bing IndexNow guide.
5. Confirm what Google saw. Check that the page is indexed and that the stored version reflects your change, then re-check anything that's still invisible a few days later.

Steps 3 to 5 are tedious at scale, which is the job FreshlyCited does for Google, whose AI Overviews and AI Mode can only link to the version in its index. It checks which of your URLs Google has indexed, sends the ones it hasn't, re-checks them on days 1, 3 and 7 and shows you the before and after. For sites you verify with an IndexNow key file, the same URLs are also sent to Bing through IndexNow in the same step, free.

Updated pages Google already has. The pre-flight check skips them by default, so no credit is used. To ask for a re-crawl anyway, choose Refresh anyway: the page is sent and uses 1 index credit when it's sent, with no credit back, because we can't see when Google re-crawls a page it already has. IndexNow notifications to Bing (for verified sites) stay free.

You can push updated pages to Google the day they ship. Being crawled sooner makes an update eligible sooner; whether it's cited is the engine's call, based on quality, authority and relevance. What you can control is asking search engines to fetch the version you worked on. For the wider strategy, see our AI search optimization guide. Check your own links free.