Does content freshness affect AI citations, and what counts as fresh?
I read Google's and Bing's documentation, four citation studies from SEO firms and three research papers, and put what they say about freshness in plain language.
On this page
The short answer
Somewhat, and mostly through real updates. Ahrefs found that AI assistants cite pages a little newer than Google's results, and Seer Interactive found 75% of cited pages had been updated within the past year. But relevance comes first, freshness matters most for time-sensitive questions, and Google and Bing say a date should change only when the content does. That is what I found after reading the sources below.
Key takeaways
- Seer Interactive found 72% of pages cited by ChatGPT, Gemini and Perplexity were under a year old by last update, but only 42% by publish date.
- Ahrefs found pages cited by AI assistants averaged 1,064 days old, against 1,432 days for Google's organic results.
- Google says its freshness systems apply where a searcher would expect fresh content, such as news, not to every question.
- Google and Bing say to change a page's date only after a real change. Lab tests show AI models react to dates alone.
- Relevance still decides first: Ahrefs found the freshest pages were often discarded within a ChatGPT answer, except for news.
- Bing's AI Performance report counts citations in Copilot, and Search Console shows impressions in AI Overviews and AI Mode.
What do Google and Bing say about content freshness?
Freshness counts where a searcher expects it, and a date should follow a real change. I read Google's ranking and date documentation and a Bing post on sitemaps in AI search.
Softlixx sells SEO and AI search work, so I have a stake here: “refresh your content for AI” is an easy service to sell. The evidence is more mixed than that.
Google's guide to its ranking systems, last updated in December 2025, describes “query deserves freshness” systems designed to show fresher content for queries where it would be expected. Its example: people searching for a newly released film probably want recent reviews rather than older articles from when production began. Google's guide to generative AI search says its AI features rely on its core Search ranking systems to retrieve relevant, up-to-date web pages, to improve the freshness of AI responses.
Bing's July 2025 post on sitemaps in AI search calls the lastmod field in a sitemap, the date a site gives for each page's last change, a key signal for deciding which pages to crawl again. Bing says it may skip a page entirely if the content has not changed since the last crawl. For how long crawling itself takes, see the research on how long SEO takes.
For AI powered search engines like Bing, freshness signals directly influence how quickly updates are reflected in search results and AI generated answers.
My readingI read this as two jobs for freshness. In Google, it is a ranking signal for some questions, not all. In Bing, it mainly decides how fast a real change reaches the answers.
Do AI assistants cite newer pages than Google search?
A little newer on average, but still years old. The two largest studies I found come from SEO firms, and they disagree in places.
Ahrefs' study, published on 28 July 2025, analysed 17 million citations across 7 AI search platforms, dating each page by when Ahrefs' crawler first saw it. Pages cited by AI assistants averaged 1,064 days old, against 1,432 days for Google's organic results, which Ahrefs calls 25.7% fresher. Google's AI Overviews leaned oldest, with the average cited article over 1,432 days old. Ahrefs adds that the average cited page is still 2.9 years old.
Average age of cited pages, in days since Ahrefs’ crawler first saw them (2025)
Shorter bars mean newer pages. Bars use a scale of 0 to 1,500 days. Google AI Overviews averaged over 1,432 days.
Source: Ahrefs.
Seer Interactive's 2025 study looked at more than 5,000 cited URLs whose publish dates it could read. About 31% of ChatGPT's citations were from 2025, about 50% of Perplexity's and roughly 44% of AI Overviews'. Seer adds that this was not a steadfast rule: behaviour varies across industries. Ahrefs sells SEO tools and Seer Interactive sells SEO and GEO services, so both have a stake in this topic too.
My readingThe two studies disagree on AI Overviews and Perplexity. They measure differently: Ahrefs averages days since its crawler saw a page, and Seer counts the share from each year on a smaller sample. I take both as “somewhat newer than search”, nothing more.
Does the publish date or the last update count?
The last update, in the most recent study I found. Old pages that are kept current get cited.
Seer Interactive's 2026 study followed ChatGPT, Gemini and Perplexity from March through June 2026. It dated 7,683 cited pages carrying 47,097 citations, taking each page's last-modified date from schema, sitemaps and headers. 75% of those pages had been updated within the past year. On the 4,124 pages where both dates were readable, 72% were less than a year old by last update, but only 42% by publish date. Gemini's citations were the most recently updated (78% within a year), then ChatGPT's (73%) and Perplexity's (65%).
Cited pages less than a year old, measured two ways
Bars use a scale of 0 to 100%. 4,124 cited pages where both dates were readable, March to June 2026.
Cited pages updated within the past year, by engine
Bars use a scale of 0 to 100%. Seer Interactive, March to June 2026.
Source: Seer Interactive.
Newest gets you the spike, established plus maintained gets you the staying power.
My readingSeer's dates come partly from sitemaps and headers, which a website can stamp automatically. So “updated” may not always mean the words changed. I take the gap as a strong hint to maintain good pages, not proof that updating causes citations.
Does it depend on the question?
Yes. Freshness matters most for news, prices and anything that changes, and much less for a stable definition.
Google's ranking guide gives an example: a search for “earthquake” usually brings preparation material, but if an earthquake happened recently, news articles and fresher content might appear. Ahrefs' 2026 study of 1.4 million ChatGPT prompts found the same split. Among the pages ChatGPT retrieved for one prompt, the older, more established pages tended to get cited and the freshest content tended to get discarded. For news queries, younger pages had a clear advantage.
A 2026 survey of 45 studies by Olivier Martinez, a preprint, rates recency, prices and dates as moderate evidence: useful but non-universal signals for time-sensitive or commercial queries. It also warns that several prominent measurement studies have not been peer-reviewed.
A recent date helps with a time-sensitive query, but not necessarily with a stable definition.
Can I just change the date on my pages?
No. Google and Bing both say the date should follow a real change, even though lab tests show AI models react to dates alone.
The lab evidence is real. A 2025 preprint added made-up publication dates to passages and asked seven language models to rank them. Fresh dates moved the top 10's average publication year forward by up to 4.78 years and moved single items by as many as 95 places. Larger models reduced the effect, but none removed it. A 2026 lab study by three Sprinklr staff, across six models and 252,000 trials, found a recent timestamp helped consistently against content dated 2019, and lists “update timestamps” among its quick wins.
The search engines say otherwise. Google's page on helpful content lists changing dates to seem fresh, without a substantial change, as a warning sign, and its page on byline dates says it does not depend on a single date factor. Its sitemap guide says lastmod should reflect the last significant update, and a new copyright date is not one. Bing says the same about lastmod.
If an article has been substantially changed, it can make sense to give it a fresh date and time.
| Source | What it says | Scope |
|---|---|---|
| Google, helpful content | Changing the date to seem fresh, without a substantial change, is a warning sign. | A self-check for site owners. |
| Google, sitemaps | lastmod should reflect the last significant update. A new copyright date is not one. | Sitemaps sent to Google. |
| John Mueller, Google, 2019 | A fresh date makes sense when an article has been substantially changed. | Visible dates. An older post. |
| Bing, 2025 | Do not set lastmod to the time the sitemap was generated unless the content changed. | Bing search and its AI answers. |
| Sprinklr lab study, 2026 | Lists “update timestamps” among its quick wins. | Six models, with no search engine. |
Sources: Google Search Central (helpful content), Google Search Central (building sitemaps), Google Search Central Blog, Bing Webmaster Blog, arXiv.
My readingI read the lab results as the reason the trick is tempting, and the search engines' rules as the reason it is risky. The lab tests hand pages straight to a model. Google says it looks at several factors to judge a page's date, and Bing may skip a page whose content has not changed.
Will fresh content get cited on its own?
No. Relevance comes first in every study I read, and freshness mostly helps when pages are otherwise close.
Ahrefs' 2026 study says it plainly: relevance still does the heavy lifting. The median page ChatGPT cited from search was around 500 days old. In the Sprinklr lab study, a recent date against an old one had a large effect in all six models, but a recent date against no date reached significance in only two or three of them. The authors add that real systems retrieve more pages than their test did.
Low-quality, irrelevant content that's updated every day will not have a magic positive effect on your visibility in AI search.
How do you know whether AI systems cite you?
Use the free reports from Bing and Google, then check your key questions by hand each month.
Bing's AI Performance report, introduced in February 2026, shows how often your content is cited in generative answers across Microsoft Copilot, AI summaries in Bing and select partner integrations. Its Total Citations figure counts the times your pages are shown as sources. Bing also says regular updates help ensure AI systems reference the most current version of your content. Google's generative AI reports in Search Console show impressions in AI Overviews and AI Mode, with a Pages view of which URLs appeared.
Bing counts citations and Google counts impressions. Neither covers ChatGPT, Gemini or Perplexity, so those need a manual check. The wider question of what earns a citation is covered in the research on getting cited by AI search. Recency comes up in local search too, mainly through recent reviews, covered in the research on Google Business Profile and local search.
My readingI would pick the 10 questions customers really ask, check each AI tool monthly, and note whether my page was cited and when it was last really updated. A few months of that shows whether updates help.
What should a business do first?
If I were starting today, I would fix the pages that go out of date first, then show their dates honestly. It is my reading of the sources above.
- Start with pages that go staleGoogle applies freshness where searchers expect it, such as news or prices.
- Update the facts, then the dateGoogle says a fresh date makes sense when the content has been substantially changed.
- Show a visible last-updated dateGoogle says to show the date and add datePublished and dateModified.
- Keep sitemap dates honestGoogle and Bing both say lastmod should reflect a real update.
- Maintain your best older pagesSeer found 72% of cited pages were fresh by last update, but only 42% by publish date.
- Keep relevance firstAhrefs found relevance still does the heavy lifting when ChatGPT picks what to cite.
- Check citations every monthBing’s AI Performance report and Search Console show where you appear.
Questions people ask
Does updating content help you get cited by ChatGPT?
Possibly, if the update is real. Seer Interactive found 73% of pages ChatGPT cited had been updated within a year. But Ahrefs found that within one answer, ChatGPT often cited older, established pages over the freshest ones, except for news. None of the studies proves that updating causes citations.
How often should I update my content for AI search?
I found no source that sets a schedule. Ahrefs warns that low-quality content updated every day will not help, and Google ties a new date to a substantial change. The closest marker I found is Seer Interactive's: 75% of the pages AI tools cited had been updated within the past year.
Should I change the published date when I update a page?
Only after a real change. Google's John Mueller wrote that a fresh date makes sense when an article has been substantially changed, and Google lists changing dates to seem fresh as a warning sign. Google suggests showing a visible date and using datePublished and dateModified in structured data.
Do Google AI Overviews prefer fresh content?
The studies disagree. Ahrefs found AI Overview citations averaged over 1,432 days old, the oldest of the systems it measured. Seer Interactive found about 44% of AI Overview citations in its 2025 sample came from that year. Google says its AI features use its core ranking systems, which favour fresher content where a query expects it.
Sources
- A guide to Google Search ranking systemsGoogle Search Central. Last updated 10 December 2025.
- Optimizing your website for generative AI features on Google SearchGoogle Search Central. Updated 10 July 2026.
- Keeping Content Discoverable with Sitemaps in AI Powered SearchBing Webmaster Blog (Fabrice Canel and Krishna Madhavan). Published 31 July 2025.
- New Study: AI Assistants Prefer to Cite “Fresher” Content (17 Million Citations Analyzed)Ahrefs (Ryan Law). Published 28 July 2025.
- Study: AI Brand Visibility and Content RecencySeer Interactive (Sonny Vasquez). Published 25 June 2025.
- Study: Content Recency’s Impact on AI Visibility in 2026Seer Interactive (Sonny Vasquez). Published 24 July 2026.
- Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts)Ahrefs (Louise Linehan). Published 15 April 2026.
- Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026)arXiv preprint (Olivier Martinez). Published 15 July 2026.
- Do Large Language Models Favor Recent Content? A Study on Recency Bias in LLM-Based RerankingarXiv preprint (Hanpei Fang, Sijie Tao, Nuo Chen, Kai-Xin Chang, Tetsuya Sakai). Submitted 14 September 2025, revised 21 September 2026.
- What Gets Cited: Competitive GEO in AI Answer EnginesarXiv, SIGIR 2026 (Rahul Vishwakarma, Shushant Kumar, Ratnesh Jamidar). Published 25 May 2026.
- Creating helpful, reliable, people-first contentGoogle Search Central. Last updated 1 October 2026.
- Influence your byline dates in Google SearchGoogle Search Central. Last updated 10 December 2025.
- Build and Submit a SitemapGoogle Search Central. Last updated 8 July 2026.
- Help Google Search know the best date for your web pageGoogle Search Central Blog (John Mueller). Published 11 March 2019.
- Introducing AI Performance in Bing Webmaster Tools (public preview)Bing Webmaster Blog. 10 February 2026.
- Introducing Search Generative AI performance reports in Search ConsoleGoogle Search Central Blog. 3 June 2026.
How I checked this. I opened every source linked here and checked each figure and each quotation against the source page, most recently on 3 October 2026. Anything in a box marked “My reading” is my own interpretation, not a finding of the studies. Studies change and new ones appear, so I will update this piece and change its dates when the evidence does.