How should we refresh old content, and how often?
Our answer
Review pages at least quarterly. Update the ones losing clicks, ranking around positions 8 to 20 or out of date. Make real changes on the same URL, then change the date.
- Confidence
- Medium
- Strongest proof
- 1 controlled test
- 4 official rules, 6 datasets, 7 practitioners, 7 opinions or claims
- Sources
- 19 sources
- 25 statements
- GRSMA
TSD
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentA new date alone doesn't make a page fresh. Google estimates when a page was significantly updated from several signals, and it lists new dates on pages that have not substantially changed as a warning sign.
Updating pages that already rank pays. Refreshed pages grew Google pageviews 28% against 9% for matched untouched ones across 5,500 pairs (Raptive, publisher ad network). Distribb (SEO content automation tool) advises updating a page ranking 8 to 20 over a new post.
AI answers cite updated old pages, not only new ones. Seer Interactive (SEO and paid media agency) found more than a quarter of the fresh pages cited were first published over two years ago. Chen and Koudas found somewhat fewer AI citations for older pages.
Our judgment
- Decide what to refresh from performance and accuracy, not a fixed calendar. Conductor says performance decides, and Google treats a new date on an unchanged page as a warning sign.
- Review at least quarterly and start with pages around positions 8 to 20 or losing clicks. We draw this from Ten Speed's quarterly review and its 4 to 20 position range, and from Distribb's view that updating a page at 8 to 20 beats a new post.
- Change the visible date and dateModified only after a substantive update on the same URL. This follows Google's date and helpful content guidance, Conductor's advice not to just change the date and Ten Speed's advice to keep the URL.
- Check pages that recommend products or state prices, models or years most often. Kevin Indig found freshness the only predictor of top three rankings for listicles, and Fang and co-authors found in tests that AI rerankers promote newer dated passages. Jacques and co-authors show older pages are still cited.
- Treat clicks, links and AI citations as separate reasons to keep an old post. Rewrite or redirect a page with any of them and delete only pages with none. We combine LSEO's rule to delete only pages with no traffic, links or role, Siege Media's retirement thresholds, which include no AI citations, TheGuideX's refusal to delete any page with links and Google's view of deletion as a last resort.
Our judgment, built from Conductor, Google, Ten Speed, Distribb, Kevin Indig, Fang and co-authors, Jacques and co-authors, Siege Media, TheGuideX and LSEO: our reading of these sources, not a finding from any one of them.
What the sources say
25 statements from 19 sourcesGoogle says AI Overviews and AI Mode have no extra requirements: a page only needs to be indexed and eligible for a snippet. Their traffic is counted in Search Console's overall search traffic.
“There are no additional requirements to appear in AI Overviews or AI Mode”
A University of Toronto study of 158,099 AI answers found older pages got somewhat fewer citations from ChatGPT and Gemini, though not consistently. It is observational and covers mainly home internet.
“Older pages receive fewer citations from ChatGPT and Gemini.”
Google lists changing page dates to look fresh without real changes, or adding and removing lots of content to seem fresh for rankings, as warning signs. It says this won't help rankings.
“Are you changing the date of pages to make them seem fresh when the content has not substantially changed?”
A Waseda University experiment found all seven AI models tested ranked passages higher when given newer dates, and a date tag alone reversed up to 25% of choices. It did not test live search engines.
“a simple date tag can reverse up to 25% of preferences between equally relevant passages.”
AirOps, an AI search software firm, says not to bulk-delete pages on traffic alone: check backlinks and seasonal trend first, and 301-redirect retired pages. Pruning alone does not add AI citations, it says.
“Avoid bulk-deleting pages on traffic alone. A page with no clicks can still hold valuable backlinks or dipped only for a seasonal reason, so confirm links and trend before you retire it.”
Kevin Indig, an organic growth advisor, says that for listicle pages in Google, only freshness predicted top-three rankings; author bios, schema and word count did not. His data is behind a paywall.
“Only freshness predicted top-three rankings.”
Conductor's Pat Reinhart says not to change a page's publish date just to look fresh for AI answers: if a page keeps earning citations, leave it be. No data is given.
“don't just go in and change the date”
On-Page.ai checked 370 page-one Google results for 50 US keywords: 46% of pages claiming an update in the past year changed only the date, and ranking showed no link to the displayed date's age. Traffic not measured.
“46% of page-one results that claim an update in the last year changed nothing but the date.”
Seer Interactive, studying 7,683 pages cited by AI tools for four client brands, found the freshness AI answers reward comes from updating old pages, not new posts. Small, correlational sample.
“The freshness LLMs reward is being manufactured by updates, not by new publishing.”
An SEO consultant who deleted only 5 of 201 old URLs says Search Console cannot show AI citations, so check pages in ChatGPT and Perplexity before removing them, and redirect when unsure.
“If a page is earning citations rather than clicks, that is an argument for rewriting it, not removing it.”
Siege Media, an SEO agency, says retire a page only with zero organic traffic for 180 days, no AI citations, falling backlinks and weak engagement, and that an AI Overview citation alone brings no visits.
“Citations don’t drive traffic.”
Raptive's study of 103,000 refreshed pages on its publisher network found refreshed pages grew Google pageviews 28% against 9% for matched untouched pages. Refreshes changing word count by over 100% averaged 44%.
“In the matched-control analysis of 5,500 refreshed and non-refreshed URL pairs, refreshed URLs averaged 28% Google pageview growth.”
LSEO, an SEO agency with an AI citation tracker, says old pages can still shape AI summaries of a brand. It redirects pages with links or demand to the closest relevant page and deletes only those with no value.
“Delete a page when it has no strategic value, no meaningful traffic, no useful backlinks, and no future role in your topical authority.”
Workshop Digital, an agency, refreshed 95 pages on its own site: sessions rose from 610 to 3,341 in the 90 days after, while untouched control pages fell 20.2%. Most wins came from focused fixes, not full rewrites.
“the biggest wins did not always come from massive rewrites”
Cyrus Shepard's review of 54 studies scores freshness 7.0 for AI citations, mattering more for recent events than established topics. He calls these correlated features, not ranking factors.
“Questions about recent events require more current information than questions about established practices.”
inblog, which sells a blog platform, says its own B2B blog's traffic rose 29% over four months with no new posts, as content takes 3 to 6 months to peak. It recommends at least 2 to 4 posts a month.
“29% increase with no new content.”
Distribb, which sells content production, advises most sites to publish two to four useful posts a month, and past 50 posts to update pages ranking 8 to 20 instead. It gives no data.
“publish as often as you can maintain the quality bar, and not once more.”
Ten Speed, which sells content refresh work to B2B software firms, says refreshes pay off fastest for pages in positions 4 to 20, publish dates should change only for substantial edits, and URLs should stay put.
“The single biggest mistake teams make is changing URLs.”
Raptive's look at winners and losers in Google's December 2025 core update found updating content tied most to gains: winning pages averaged 393 days since refresh, against 500. Sample size not given.
“updating existing content had the strongest positive relationship with performance”
An academic study of 615 sources ChatGPT cited for 100 health questions found over 75% came from established institutions, and 36.4% of cited content dated from 2024 to 2026 while 40.7% predated 2020 or had no date.
“Content Recency - Before 2020/No Date 250 (40.7%)”
Brett Tabke argues content pruning should be judged on AI citations and brand mentions as well as clicks, since pages can shape answers with no traffic, and results settle over weeks 6 to 12.
“If you only count clicks, you will systematically undervalue the pages doing the most work.”
Google says it estimates a page's byline date from several signals, and advises a prominent visible date labeled Published or Last updated, consistent with datePublished or dateModified markup.
“Add a user-visible date to the page and feature it prominently.”
Google says most sites need not react to core updates; wait a week after one finishes, compare with the week before, avoid quick fixes, and expect recovery from big drops to take months.
“We recommend waiting at least a full week after a core update completes before analyzing your site in Search Console.”
Siege Media, a content and SEO agency, refreshes quarterly the posts ranking on pages 1 to 3 untouched for 3 or more months, and reports a 70% organic traffic rise from it, with no sample or comparison given.
“Ranks on page 1-3 for target keywords (these have momentum worth amplifying).”
AirOps claims pages not updated in over three months are more than 3 times as likely to lose AI citations, and advises updating key pages every three months. It gives no sample, period or method.
“Update key pages every three months so models treat your information as current, accurate, and aligned with how people search.”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it2 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer