When an AI answer engine picks a source to quote, it is usually choosing between dozens of pages that say roughly the same thing. One of the fastest tie-breakers is recency: can the engine verify that this page is current? A page with a visible “Updated March 2026” line and a matching dateModified value in its structured data beats an undated page making identical claims — because the engine can defend quoting it.
Freshness matters more to answer engines than it ever did to classic search. A search engine ranks ten links and lets you judge them; an answer engine states one answer as fact. If it quotes a stale source, the engine itself is wrong. So retrieval systems are tuned to prefer sources whose recency they can verify. This guide covers how to send that signal honestly — and why faking it backfires.
Why do answer engines care so much about dates?
Because the engine, not the reader, carries the blame for a wrong answer. When ChatGPT, Perplexity, or Google AI Overviews assembles a response, it retrieves a handful of candidate pages and synthesizes them into a single statement. For any topic where the truth can change — prices, regulations, product features, statistics, recommended practices — a verifiable date is evidence that the page still reflects reality.
There is also a structural reason. Language models have training cutoffs, and retrieval exists precisely to fill the gap with current information. A page that cannot demonstrate its own currency defeats the purpose of retrieval. An undated page forces the engine to guess how old the information is, and engines tend to resolve that guess by preferring the competitor who dated their work.
How do AI engines verify a page's recency?
Through several overlapping signals — and agreement between them is the real test. The main ones:
- Visible dates. A human-readable “Published” and “Last updated” line near the top of the page. This is the date a reader (and a quoting engine) can actually see in the text.
- datePublished and dateModified in schema markup. Structured data — Article or BlogPosting markup using the dateModified property — gives machines an unambiguous timestamp. Our guide to schema markup for AEO covers the full implementation.
- Sitemap lastmod. The lastmod field in your XML sitemap tells crawlers which pages changed and when. It is one of the main inputs crawlers use to prioritize re-visits.
- The content itself. Years, version numbers, and prices in the copy. If your schema says 2026 but a paragraph says “as of 2023,” the copy wins — and you lose.
When these signals disagree — the visible date says one thing, the schema another, the sitemap a third — engines discount all of them. Consistency is the signal.
How do retrieval-based engines re-crawl your site?
On their own schedule, weighted by how often your site demonstrably changes. Engines like Perplexity and ChatGPT's browsing mode run their own crawlers (PerplexityBot, OAI-SearchBot), and several lean on established web indexes — ChatGPT's search draws heavily on Bing, and AI Overviews sits on Google's index. Either way, the mechanics are the same: sites that update regularly and announce it through an accurate sitemap earn more frequent re-crawls; sites that never change get visited rarely, so even a genuine update can take weeks to be noticed.
Three practical consequences. First, keep your sitemap's lastmod values honest and automatic — generated from real modification times, not hardcoded. Second, make sure AI crawlers can reach your pages at all; a blocked crawler never sees your update. Third, a steady cadence of real updates trains crawlers to come back. For the full pipeline from crawl to citation, see how AI answer engines work.
How do you update evergreen pages honestly?
Change the substance first; the date follows. An honest update workflow looks like this:
- Re-verify every factual claim on the page: numbers, prices, names, recommendations, links.
- Rewrite what changed and add what is genuinely new since publication — a new consideration, a changed regulation, a better example.
- Then bump the visible “Updated” line and the dateModified schema value, together.
- Keep datePublished stable. Resetting the publication date throws away the page's age, which is itself a trust signal.
What you should never do is bump the date without touching the content. Search engines and answer engines keep historical copies of pages; a dateModified that jumps while the content stays byte-for-byte identical is a detectable pattern, and it reads as manipulation. Google has said for years that artificially freshening dates does not help rankings — and for answer engines the stakes are higher, because a date the engine cannot reconcile with an unchanged page undermines every other signal you send. Fake freshness is one of the most common self-inflicted wounds we see in audits.
What stale claims should you prune?
Anything a reader could falsify with a calendar. When you review an evergreen page, hunt for:
- Year references — “in 2024, the best approach is…” instantly dates the whole page.
- Prices, plans, and fees that have changed since publication.
- Statistics more than a couple of years old presented as current.
- Tools, products, or features that have been renamed or discontinued.
- Words like “new,” “recently,” and “upcoming” attached to things that no longer are.
Pruning matters because answer engines quote at the sentence and passage level, not the page level. One stale sentence can be the exact passage retrieved — or the reason the engine passes over an otherwise strong page. Our guide to writing content AI can quote covers how to structure those passages once they are accurate.
Freshness vs. authority: which one wins?
It depends on how volatile the query is. For time-sensitive questions — anything involving prices, laws, versions, or “best X in 2026” — freshness dominates: a well-maintained page on a modest site routinely gets cited over a stale page on a famous one. For stable topics — how escrow works, what a root canal involves — authority dominates and the date matters much less.
The winning position is both: an authoritative source that is also verifiably maintained. And here is the realistic opportunity for smaller sites — you will not out-fresh a major publisher on breaking news, but on the evergreen commercial topics in your niche, most competitors have not touched their pages in years. Being the most recently verified source on a specific question is an edge you can actually earn.
Frequently asked questions
How often should I update evergreen content?
As often as the facts change, checked on a schedule. A quarterly review of your most important pages is a reasonable cadence for most businesses: re-verify claims, update what moved, and leave the date alone if nothing did. Updates should be fact-driven, not calendar-driven.
Does changing the date without changing the content help?
No. Engines can compare the current page against stored copies, so a new date on unchanged content is detectable — and it damages the credibility of your dates everywhere else on the site. If anything, it makes engines trust your genuine updates less.
Should I show both a published date and an updated date?
Yes. Show both in the visible copy and mirror them in schema as datePublished and dateModified. The published date proves the page's age and track record; the updated date proves it is still being maintained. Together they answer the two questions an engine is asking.
Do AI engines really read schema dates?
The retrieval systems behind them do. Most answer engines draw on web indexes (Google's, Bing's, or their own crawls) that parse structured data, and dates are among the most consumed properties. Schema alone will not earn a citation — but a missing or contradictory date is an easy reason to lose one.
Not sure whether your visible dates, schema, and sitemap are telling engines the same story? Run our free AEO audit — it checks your date signals along with the rest of your AI visibility, and if something needs fixing we handle remediation at $99/hr with ongoing monitoring from $19/mo.