Do AI assistants cite fresher content than Google? Ahrefs data

Freshness advice is usually anecdotal. Ahrefs published a large dataset that at least lets the claim be checked against something: how old are the pages that AI assistants actually cite? This write-up reports that study by Ryan Law and Xibeijia Guan, published on 28 July 2025. SEO Madman did not run it.

What did Ahrefs measure?#

Ahrefs used Brand Radar data to extract 16.975 million cited URLs from ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews and organic Google results. For each URL it measured days since publication and days since the last update, then compared averages by platform. The ai-search hub tracks related studies.

Do AI tools cite fresher pages?#

Yes, on average. URLs cited by AI assistants averaged 1,064 days old versus 1,432 days for organic results, a gap Ahrefs reports as 25.7 percent. Measured by last update instead of publication, the gap shrinks: 909 days for AI assistants against 1,047 for the organic results, which Ahrefs puts at 13.1 percent.

The platform breakdown is uneven. By days since publication, Ahrefs reports ChatGPT citations at 958 days, ChatGPT in-text references at 1,023, Copilot at 1,056, Gemini at 1,118 and Perplexity at 1,166. ChatGPT citations were 458 days newer than organic results, the strongest preference in the set. Google’s top three AI Overview citations were the exception, averaging 1,432 days and about 16 days older than organic results.

One detail is worth checking before reusing the headline number. Ahrefs’ table lists the organic results at 1,416 days, while its 25.7 percent calculation divides by 1,432, the AI Overviews figure. The direction of the finding does not change, but the exact percentage depends on which baseline is used.

Does citation order change with age?#

For two platforms, it appears to. Ahrefs found that most platforms cite older content first, mirroring organic rankings, but that Perplexity and ChatGPT references show a positive correlation between citation order and content age, with newer content cited first. Ahrefs describes this as what looks like intentional age-related ordering, an inference from the pattern, not a documented rule.

Do new dates raise citations?#

The study cannot say. The average cited URL is still about 2.9 years old, and the study compares ages of cited pages, not before-and-after edits. Ahrefs itself notes that John Mueller has warned against changing publish dates without changing content, and that freshness is one factor among many. The authors add that most brands will likely do better creating new, high-quality content than committing resources to extremely frequent updates.

What can be acted on from this?#

Google’s own guidance draws the same line. Its helpful content questions ask whether you are “changing the date of pages to make them seem fresh when the content has not substantially changed,” and treat that as a warning sign. Its byline date documentation asks for a visible date labeled as published or last updated, datePublished or dateModified in Article or BlogPosting structured data, and for the visible and structured dates to match. It also says not to give future dates or the date of the event the page describes.

So the practical steps are about accuracy, not tactics. Review pages that must stay correct, change the modified date only when the content changes, and keep sitemap lastmod values honest; the sitemap lastmod format auditor checks their format. The study supports neither a refresh schedule nor a claim about causation.