Why Your "Last Updated" Date Decides Whether AI Cites You

Freshness is one of the strongest selection signals in AI citation — particularly for Perplexity, which can begin citing a well-structured new page within hours for trending topics, and for Google's AI Overviews, which actively factor publication and modification dates into source selection. A page updated last month beats an otherwise identical page left untouched for two years, especially for time-sensitive queries — for genuinely evergreen topics, the gap narrows considerably. The signal has to be machine-readable (dateModified in Article schema), human-readable (a visible date on the page), and accurate — updating the date without updating the content is detectable and counterproductive. Genuine freshness compounds: a page revised quarterly for three years earns more citation authority than a new page claiming the same update cadence.

How AI engines use freshness signals

AI retrieval systems process freshness at two stages. At retrieval, recency narrows the candidate pool for queries where current information matters most — "best practices in 2026," "current pricing," "latest research." An outdated page may not enter the retrieval pool at all for genuinely time-sensitive queries. At re-ranking, freshness is one of several signals that determines which retrieved candidate gets cited: between two equally relevant passages, the one with a recent dateModified tends to win.

The freshness weight varies meaningfully by engine and by query type. Perplexity weights it most aggressively for time-sensitive and trending queries — well-structured new content can earn citations within hours of publication in that context. It weighs freshness considerably less for stable, evergreen topics, where a well-maintained older page still competes fine. Google AI Overviews weighs freshness strongly for informational and research queries, consistent with the "query deserves freshness" logic that's long existed in traditional search. ChatGPT's live retrieval path weighs freshness moderately; its training-based path has no freshness signal at all, since it reflects a fixed historical snapshot rather than a live crawl.

The three freshness signals

1. Article schema dateModified. The most important machine-readable freshness signal. Update this field every time you meaningfully revise a page. AI crawlers read it directly from your JSON-LD without inferring a date from prose. If you update content but not dateModified, the crawler has no signal the page changed at all.

2. Visible "Last updated" date. Human-readable, using a <time datetime="YYYY-MM-DD"> element for machine precision alongside the visible text. This matters both for AI systems that extract dates from prose rather than schema, and for readers who use the date to judge reliability. Place it near the article title or byline — not buried in the footer.

3. Content substance. The date alone isn't enough, and this is the part AI systems are increasingly able to detect: a page whose dateModified says last month but whose facts, statistics, and recommendations are visibly two years stale is a mismatch a careful retrieval system can catch. Updating the timestamp without updating the substance is a freshness signal that doesn't actually deliver freshness.

What counts as a meaningful update

Not every revision justifies a freshness update. Changes that genuinely signal freshness to AI systems:

• New statistics replacing outdated ones, with updated source citations

• New sections addressing developments since the original publication

• Revised recommendations reflecting current best practice

• Corrected facts or removed outdated claims

Changes that don't meaningfully affect freshness: fixing typos, reformatting, swapping images, adjusting headings without revising substance underneath them.

The update cadence that works

For evergreen GEO content — how-to guides, explainers, checklists — a quarterly review cycle is a practical standard. Check each high-priority page once per quarter: are the statistics still current? Are the recommendations still accurate? Has anything in the category changed that affects the content? If yes, update the content and refresh the date. If no, leave the date alone — updating it without a substance change is exactly the kind of mismatch that undermines trust in the signal.

For fast-moving topics (AI tools, platform features, regulatory changes), monthly review is warranted. For genuinely stable reference content (definitions, historical background), semi-annual review is sufficient — and this is precisely the category where the freshness signal matters least to begin with, since it's not what Perplexity or Google are weighting heavily for a non-time-sensitive query in the first place.

Frequently asked questions

Should I update old posts or write new ones? Update high-performing old posts first. A page that already has citation authority and inbound links accumulates freshness value from a revision in a way a new URL cannot. New URLs start the citation clock from zero; existing pages compound.

Does changing the URL reset my freshness authority? Yes — and it loses all existing backlinks and citations unless you 301 redirect. Never change a URL for freshness reasons. Update the content at the existing URL and refresh dateModified.

How do I show both a publish date and an update date? Use two <time> elements — one for datePublished, one for dateModified — both with ISO 8601 datetime attributes. Show dateModified prominently ("Last updated September 2026") and the original publish date in smaller text for transparency.

Does freshness matter for every type of content? No — and this is worth being precise about. It matters most for time-sensitive and trending queries, where Perplexity in particular weighs recency heavily. For genuinely evergreen, stable topics, a well-maintained older page still competes reasonably well; don't burn review cycles chasing a freshness bump that specific query type doesn't reward.