Does changing the date get a page cited by AI? What counts as an update

KEY TAKEAWAYS

1. Fresh means recently updated. In Seer's 2026 data 72% of cited pages were fresh by update date and 42% by publish date.
2. A date label alone moved ChatGPT, Claude and Perplexity in one test. Google AI Mode and Gemini fell in the same test.
3. Change the visible date, dateModified and sitemap lastmod together, and only when content, links or structured data changed.
4. Review comparison pages monthly, product pages quarterly and guides twice a year. Update on facts, not on the calendar.

Most advice on freshness for AI search says two things. Update often, and do not bother changing the date without changing the content, because the engines can tell. The first is roughly right. The second is not what the evidence shows. This article separates what the data says about fresh pages, what a date label does on its own, and what the platforms say they check. It ends with a table of what counts as an update and a review rhythm by page type.

AI engines cite fresher pages, but not new ones

Ahrefs published the largest study in July 2025. It compared 16.975 million URLs cited by ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews with organic results. Cited URLs averaged 1,064 days since publication against 1,432 for organic results, which is 25.7% fresher. Two details rarely make the summaries. The average cited page was still 2.9 years old. And in the per-platform breakdown, AI Overviews cited pages averaging 1,432 days old against 1,416 days for organic results, 16 days older. In that dataset Google's AI layer showed no preference for newer pages.

Ahrefs returned to the question in April 2026 with 1.4 million ChatGPT prompts. Pages cited from the search index had a median age of about 500 days. Its conclusion was that relevance does most of the work and freshness decides mainly among news results.

Seer Interactive's study of July 24, 2026 explains the pattern. It dated 7,683 pages that carried 47,097 citations between March and June 2026 in ChatGPT, Gemini and Perplexity, using last-modified dates from schema, sitemaps and headers. Three quarters of those pages had been updated in the last year: 78% for Gemini, 73% for ChatGPT, 65% for Perplexity.

Then Seer measured the same pages two ways. Of 4,124 pages with both dates readable, 72% were fresh by update date. By publish date, 42% were. The freshness the engines reward mostly sits on older URLs that somebody maintains. The data comes from four client categories, with retail energy over-represented, so treat the figures as indicative.

Perplexity's developer documentation makes the same distinction in its search filters. One set filters by publication date. A separate set filters by when a page was last updated.

A date label alone can move the result

A paper from researchers at Waseda University and Hong Kong Polytechnic University, posted to arXiv in September 2025, tested this in the lab. They added one line, a publication date, to the front of each passage and changed nothing else. Seven language models then reranked the passages. Newer-dated passages rose. The average year of the top ten moved forward by up to 4.78 years, single passages moved by as many as 95 places, and the choice between two equally relevant passages flipped up to 25% of the time on average.

OtterlyAI ran the field version and published it on September 24, 2026. On August 4 it changed the visible "Last updated" date on 10 of its own articles and left the content alone. Across 320 tracked prompts on seven engines, those articles drew 4,800 citations in the 30 days before and 6,312 in the 30 days after, a rise of 31.5%. Two untouched articles fell from 358 to 320.

So the claim that a cosmetic date change does nothing is too strong. That is not a reason to do it.

Why the date-only result is a poor bet

Split by engine, the result looks different. Otterly reports that ChatGPT went from 37 to 293 citations and Claude from 1,135 to 2,118. Perplexity rose 217% and AI Overviews 15%. Google AI Mode fell 20% and Gemini fell 37%. By our calculation, ChatGPT and Claude added 1,239 of the 1,512 net citations gained, or 82%. Claude alone added 65%. One engine made most of the headline.

The ChatGPT figure has a second problem. Otterly's own tracking, published on August 27, shows ChatGPT changing its sources in two steps on August 8 and August 14. Reddit citations fell 73.4% across 16 brand reports, and official brand and competitor pages rose in 10 of 15 reports. The test articles were pages on a brand's own site, and the measurement window opened on August 5. Otterly's write-up does not connect the two. We think the overlap matters. Otterly does describe its result as directional, and its earlier test of adding the year to titles ended without a result: citations rose 56% and 61% in two groups, while a page nobody touched rose 63%.

The platforms' rules point the same way. Google's sitemap documentation says it uses lastmod only when the value is consistently and verifiably accurate, and that a changed copyright date is not a significant update. Bing's webmaster team said in July 2025 that lastmod should reflect the true modification time of the content, and that freshness signals affect how quickly changes reach AI answers. Google's guide to its generative AI features says they retrieve pages through its core Search ranking systems. A date that Google cannot verify is unlikely to help in AI Overviews or AI Mode. That is our inference, but it fits the engine split above.

A date-only change may lift citations on engines that take the label at face value. It does nothing reliable on Google's surfaces, and it spends the trust that makes your real update dates useful.

What counts as an update

Google's byline date guidance adds two points. Show a visible date and label it as published or last updated. Use datePublished and dateModified in structured data. Our addition is to keep the original publish date when you update.

How often should you update?

The guides disagree. AuthorityTech recommends monthly for ChatGPT, quarterly for Perplexity and Gemini, and yearly for AI Overviews. ALF Design Group recommends every 90 days. Foglift argues there is no universal interval. The multipliers behind the fixed schedules are hard to check. A 3.2 times lift for pages updated within 30 days and a claim that half of AI citations are under 13 weeks old both circulate widely. In the articles we opened, each was credited to another firm with no link or method, so we have left them out. AirOps reports that pages left more than three months without an update are over three times more likely to lose visibility, but its report page does not describe the sample.

Two findings with a stated method are more useful. Scrunch, working with Stacker, tracked 3.5 million citation events from September 2025 to March 2026. Half of a group of citations had gone after about 4.5 weeks: 3.4 weeks on ChatGPT, 5.8 on Perplexity. That measures how fast engines rotate sources. Scrunch does not claim an update resets the clock.

Seer measured the pages that resist rotation. Pages cited in only one of four months were 86% fresh, with a median update age of 0.16 years. Pages cited in all four months were 68% fresh, with a median of 0.47 years, or about five and a half months. In the author's words: "Newest gets you the spike, established plus maintained gets you the staying power."

That supports fact-triggered updates on a review rhythm. The rhythm below is our recommendation, not a platform rule.

A refresh method in six steps

1. List the pages that matter. Start with pages already cited for your target questions, then pages that should be. Measuring AI visibility covers how to find both.
2. Check facts, not prose. At each review, test every statistic, price, product name, screenshot, step and outbound link against its source.
3. Change the date only when the table says yes. Set the visible date, dateModified and the sitemap lastmod to the same day. Keep datePublished and the URL.
4. Leave unchanged pages alone. A review that finds nothing is a good result.
5. Make sure the page gets fetched again. The GEO Lab's log of one page, published on August 31, 2026, found citation followed quickly once the crawler returned. It is a single observation, but it matches Bing's advice to submit changed URLs through IndexNow. Check that AI crawlers can reach the page.
6. Measure with a control. Compare 30 days before and after against a set of untouched pages, and note any engine change in the window.

Where freshness sits in digital authority

Freshness protects citations you have earned. It does not create the trust that earns them. In digital authority terms, updating your own pages maintains search presence and AI visibility. It does nothing for press coverage, because you cannot edit a publication's article about you. Seer found news and editorial pages were the oldest content cited, with only 45% updated in the last year. Coverage ages too, and the only way to refresh it is new coverage in the sources the engines cite. That is what the placement marketplace is for.

Questions

How often should I update content for AI citations?

Update when a fact changes, and review on a rhythm set by page type. Monthly for pricing and comparison pages, quarterly for product pages, twice a year for guides. In Seer Interactive's 2026 data, pages cited every month had a median update age of about half a year.

Does changing only the publish date increase AI citations?

It can, unevenly. Otterly changed only the visible date on 10 articles in August 2026 and citations rose 31.5%. Two engines produced most of the gain, Google AI Mode and Gemini fell, and ChatGPT changed its sources during the test. Google says it uses sitemap dates only when they are verifiably accurate.

What counts as a significant update?

Google's sitemap documentation names changes to the main content, the structured data or the links on a page. It says a new copyright date does not count. Replacing an outdated statistic, price or step qualifies. Rewording a sentence or swapping an image does not.

Is new content better than updated content for AI visibility?

Not on this evidence. Seer found 72% of cited pages looked fresh by update date and only 42% by publish date. Most fresh cited pages are older URLs that someone maintains. Keep the URL and update it.

Sources

1. Ahrefs, New Study, AI Assistants Prefer to Cite Fresher Content (17 Million Citations Analyzed) ahrefs.com
2. Seer Interactive, Study, Content Recency's Impact on AI Visibility in 2026 seerinteractive.com
3. Ahrefs, Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts) ahrefs.com
4. OtterlyAI, We Updated Only the Publication Date on 10 Articles. AI Citations Rose 31.5% otterly.ai
5. OtterlyAI, ChatGPT cut Reddit citations by at least 73% in August 2026 otterly.ai
6. OtterlyAI, Year in Title, Does It Increase AI Citations? GEO Experiment otterly.ai
7. arXiv (Fang, Tao, Chen, Chang, Sakai), Do Large Language Models Favor Recent Content? A Study on Recency Bias in LLM-Based Reranking arxiv.org
8. Google Search Central, Build and submit a sitemap developers.google.com
9. Google Search Central, Add a byline date to Google Search results developers.google.com
10. Google Search Central, Optimizing your website for generative AI features on Google Search developers.google.com
11. Microsoft Bing Webmaster Blog, Keeping Content Discoverable with Sitemaps in AI Powered Search blogs.bing.com
12. Perplexity, Search Date and Time Filters docs.perplexity.ai
13. Scrunch, The half-life of AI citations scrunch.com
14. AirOps, The 2026 State of AI Search airops.com
15. The GEO Lab, AI Citation Lag, How Long Before AI Search Cites a New Page? thegeolab.net
16. AuthorityTech, How Often to Refresh Content for AI Citations authoritytech.io
17. ALF Design Group, Content Refresh Cadence, The 4.8x AI Citation Rule alfdesigngroup.com
18. Foglift, Content Freshness in AI Search, Refresh Checklist foglift.io
19. Apiserpent, Freshness Wins, Recent Content Earns 3x More ChatGPT Citations apiserpent.com
20. Salespeak, Content Freshness and AI Search, Why 50% of AI Citations Are Under 13 Weeks Old salespeak.ai

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →