How Old Are AI Search Citations? Source Age Data 2026

Research into AI search citations consistently shows that Perplexity favours recently published or updated content, with a significant share of its cited sources published within the preceding 12 months. ChatGPT's browsing mode and Google AI Overviews also show recency bias, though the pattern varies by query type. For UK businesses, this means content staleness is a measurable risk to AI visibility.

The core statistic

A study by Seer Interactive, published in early 2026 and covering more than 4,000 AI-generated responses across ChatGPT, Perplexity, and Google AI Overviews, found that approximately 65% of URLs cited by Perplexity had been published or substantially updated within the 12 months prior to the query date. For Google AI Overviews the figure was closer to 52%, and for ChatGPT's browsing-enabled responses it was around 47%.

These figures place recency alongside authority and topical relevance as one of the primary signals influencing which sources AI platforms select when constructing a response.

Methodology and what the research measured

The Seer Interactive study sampled queries across four broad categories: informational, commercial, navigational, and local. Researchers recorded the URLs appearing in AI-generated citations, then checked each URL's published or last-modified date using a combination of HTTP headers, structured data timestamps, and Wayback Machine records where needed.

Source age was defined as the time between the URL's most recent substantial content update and the date the query was run. Minor edits such as correcting a typo were excluded. The researchers focused on changes that altered at least a paragraph of substantive information.

It is worth noting that the study was conducted between October 2025 and March 2026, a period during which several AI platforms updated their retrieval models. Results may not reflect behaviour from a single static model version.

Platform-by-platform breakdown

Source: Seer Interactive AI Citation Study, Q1 2026. Sample size: 4,000+ AI-generated responses.

Does query type change the picture?

Yes, substantially. The recency bias is strongest for news-adjacent and commercial queries, where all three platforms showed over 70% of citations coming from content under 12 months old. For more evergreen informational queries, such as definitions or explanations of stable concepts, older content performed comparably to newer material.

This matters for UK businesses deciding where to focus update efforts. A law firm's page explaining what a leasehold is may not need quarterly refreshes. A page covering current mortgage rates or recent regulatory changes almost certainly does.

Why recency matters to AI retrieval systems

Perplexity uses a live web index and explicitly signals to users that its answers are current. Its retrieval model therefore weights recently crawled pages more heavily than platforms relying on older training data snapshots. Google AI Overviews draws from the same index used for Google Search, where freshness has long been a documented ranking signal for certain query types.

ChatGPT with browsing enabled operates differently. It fetches pages at query time rather than relying on training data, but its selection of which pages to fetch still reflects an underlying preference for sources that appear active and authoritative. A page last updated in 2021 with no subsequent signals of activity is less likely to be retrieved, even if its content remains accurate.

### How Old Are AI Search Citations? Source Age Data 2026: key figures

| Metric | Value | Source |
| --- | --- | --- |
| Share of Perplexity citations from content published or updated within the preceding 12 months | 65% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of Google AI Overviews citations from content published or updated within the preceding 12 months | 52% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of ChatGPT browsing-enabled citations from content published or updated within the preceding 12 months | 47% | Seer Interactive AI Citation Study, Q1 2026 |
| Sample size of AI-generated responses studied | 4,000+ | Seer Interactive AI Citation Study, Q1 2026 |
| Share of Perplexity citations from content 1-3 years old | 24% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of Perplexity citations from content over 3 years old | 11% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of Google AI Overviews citations from content 1-3 years old | 31% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of Google AI Overviews citations from content over 3 years old | 17% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of ChatGPT browsing-enabled citations from content 1-3 years old | 33% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of ChatGPT browsing-enabled citations from content over 3 years old | 20% | Seer Interactive AI Citation Study, Q1 2026 |
| Share of citations from content under 12 months old for news-adjacent and commercial queries across all three platforms | over 70% | Seer Interactive AI Citation Study, Q1 2026 |

Source: Seer Interactive AI Citation Study, Q1 2026

What these numbers mean for your business

The data points to a clear practical implication: content that has not been updated recently is at a structural disadvantage in AI search, particularly on Perplexity. This is not simply about adding new information. It is about signalling to AI crawlers that a page reflects a current understanding of its topic.

For UK businesses, several content types carry the highest staleness risk:

• Service pages that quote prices or reference regulations
• Comparison or review pages where the competitive landscape has shifted
• FAQ pages that have not been revisited since original publication
• Blog posts citing statistics from before 2024

The risk is not just invisibility. Outdated content that does get cited can lead to AI platforms presenting inaccurate information under your brand, which creates a separate reputation problem. You can read more about that risk in our guide on preventing AI platforms from using outdated information.

How to act on this data

Audit your content by last-updated date

Run a crawl of your site and filter pages by their published or last-modified date. Any substantive page not updated in the past 12 months should be reviewed. Prioritise pages that target commercial or news-adjacent queries first.

Use accurate timestamps in structured data

Make sure your pages include dateModified in their structured data, and that this timestamp reflects genuine content changes rather than automated refreshes. AI crawlers and Google's systems both reference this field. Our schema and AI guide covers implementation in more detail.

Update substantively, not superficially

Changing a date in a footer or adding a sentence does not constitute a substantial update. The Seer Interactive methodology excluded these changes, and AI retrieval systems are increasingly capable of distinguishing meaningful content changes from cosmetic ones.

Build a content refresh calendar

Rather than updating pages reactively, schedule quarterly reviews of your highest-priority pages. This creates a consistent freshness signal and reduces the risk of large portions of your site becoming stale simultaneously.

Comparison: recency vs other citation signals

Recency is one factor among several. The same Seer Interactive study found that domain authority remained the strongest single predictor of citation across all three platforms. Pages on high-authority domains were cited at higher rates regardless of age. Recency matters most when authority is roughly equal between competing sources.

Topical relevance, use of structured data, and crawlability also featured as significant predictors. A recently updated page on a low-authority domain with poor structured data will still lose to an older but authoritative and well-structured source in most cases.

For a fuller picture of which signals matter most across platforms, see our AI search statistics hub.

Frequently asked questions

Does updating old content guarantee better AI citation rates?

Not by itself, but it removes a structural disadvantage. Content that is genuinely updated with relevant, accurate information on a high-authority domain is well-positioned for citations. Updating content on a low-authority or poorly structured site will have limited effect on its own.

How does Perplexity's recency preference compare to Google AI Overviews?

Perplexity shows a stronger recency preference, with 65% of its citations coming from content under 12 months old compared to 52% for Google AI Overviews. Perplexity's live-index architecture explains much of this difference, as it is designed to surface current information rather than evergreen content.

Should I add a "last updated" date to every page?

Visible last-updated dates help users and can reinforce freshness signals when combined with accurate dateModified structured data. On their own, visible dates without genuine content changes are unlikely to influence AI citation behaviour meaningfully.

Are there types of content where recency matters less for AI citation?

Yes. Evergreen definitional content, stable how-to guides, and foundational explainers show less recency bias in the data. For queries about fixed concepts or processes that do not change frequently, older high-quality content can compete effectively with newer material.

If you are unsure how your current content is performing across AI platforms, a free AI visibility audit can identify which pages are most at risk from content staleness and other citation barriers.