Does Content Freshness Matter in AI Search? What 4 Million Citations Show

Does Content Freshness Matter in AI Search? What 4 Million Citations Show

Does Content Freshness Matter in AI Search? What 4 Million Citations Show

Escrito por:

SEO & GEO Specialist @ aiclicks.io

Revisado por:

Pragati Gupta

Content Marketing Manager @ aiclicks.io

Última actualización:

Verificado por expertos

No headings found on page

Llegue a millones de consumidores que están usando IA para descubrir nuevos productos y marcas

Llegue a millones de consumidores que están usando IA para descubrir nuevos productos y marcas

Key Takeaways


  • 91.3% of dated citations go to pages updated within 12 months; 33.1% to pages updated within 30 days.

  • The update date predicts citations better than the publication date (−0.187 vs −0.157), but both links are weak.

  • Updated pages beat never-updated pages in all six age bands; over two years old, they average twice the citations (7.9 vs 3.9).

  • ChatGPT cites the freshest pages (median 73 days since update); AI Overviews the oldest (175 days).

  • 32.1% of all citations go to pages with no date, including 52.0% of citations to blog and news articles.

  • Old content still gets cited: 18.5% of dated citations go to pages first published over two years ago, mostly listicles and articles.

Keep your content fresh” is some of the most repeated advice in AI search, and it rarely comes with a number attached. That leaves content teams guessing whether to refresh every quarter, change the date and hope, or leave a page alone while it’s still getting cited.

We wanted an actual answer. So we crawled the 527,656 pages that ChatGPT, Gemini and Google AI Overviews cited in one week, read the publication and update date on each, and set those dates against 4 million citations.

The answer is yes, with a catch. 91.3% of citations to dated pages went to pages updated within the past year, and updated pages out-cited never-updated pages of the same age in every age band. But a recent date on its own predicts very little, and pages carrying automatic date stamps showed almost no link with citations.

Below, we break down how much a refresh is worth, which engine rewards recency most, and which of your pages to update first.

Columns: 91.3% of citations to dated pages go to pages updated within the last year; 33.1% within 30 days

Figure 1. Share of citations to dated pages, by time since the page was last updated. 91.3% go to pages updated within a year. Dated pages only (49.4% of all citations).

Show the data

Time since last update

Citations

Share of dated citations

0–30 days

656,047

33.1%

31–90 days

494,803

24.9%

91–180 days

388,887

19.6%

181–365 days

272,172

13.7%

1–2 years

92,800

4.7%

Over 2 years

80,257

4.0%

About this data. 527,656 cited pages crawled · two dates read per page · ChatGPT, Gemini and Google AI Overviews · 49.4% of citations have a usable date. The citations come from our study of which sources AI engines cite; the crawl method is below.

Most advice on content freshness comes in one of two forms: a claim that fresh content is critical for AI search, or a list of refresh tips with no numbers behind them. We wanted to know how much freshness is worth, so we read the dates on every page the three engines cited and set them against how often each page was cited. Freshness matters, and the last update counts for more than the publication date. It matters far less than the "critical" framing suggests, though, and changing a date without changing the page buys nothing we could measure.

How did we measure content age?

We took the 527,656 URLs that ChatGPT, Gemini and Google AI Overviews cited across 535,560 answers (20–29 October 2026) and crawled each one between 30 October and 4 November. For every page we read two dates: when it was first published and when it was last updated. Age is the citation date minus the publication date; time since update is the citation date minus the update date.

Our crawler looked for dates in five places, in this order, and took the first hit for each date: structured data on the page (the datePublished and dateModified properties), meta tags, microdata, a <time> tag in the article or header, and finally a date in the URL path. Three platforms got their own handling because their pages don't carry a reliable date in the HTML. Reddit posts take their creation time, YouTube videos their upload date, and Wikipedia pages their first and latest revision from the MediaWiki API. Reddit and YouTube show no edit date, so for them the update date equals the publication date.

The crawler kept at most one open request per site and waited at least two seconds between requests to the same site. Before the full run, it passed 50 offline test cases and 47 hand-picked real URLs covering news, blogs, listicles, reviews, product pages, PDFs, missing pages and bot-protected sites.

Publisher dates aren't always real, so we flagged three kinds of suspect date and kept them out of every age figure: sites that stamp one date on 80% or more of their pages (or show the crawl date as the update date), update dates earlier than publication dates, and dates in the future. 13,736 pages carried at least one flag.

Every cited page then falls into one of five groups. Only the first group, dated pages, feeds the age figures in this post, and it holds 49.4% of all citations. The "citation date" is the last day in the window that an engine cited the page.

Half of AI citations (49.4%) go to pages with a usable date; 32.1% to undated pages, 12.1% to pages that couldn't be fetched

Figure 2. All 4,018,882 citations by what the crawler found on the cited page. 49.4% go to dated pages, the group every age figure in this post uses.

Show the data

Group

Pages

Citations

Share of all citations

Dated

218,795

1,984,966

49.4%

Flagged (suspect dates)

12,206

145,788

3.6%

Date after citation

10,029

113,195

2.8%

Undated

202,495

1,290,426

32.1%

Not fetched

84,131

484,507

12.1%

Do AI engines cite fresh content more?

Mostly, yes. 91.3% of citations to dated pages go to pages updated in the last 12 months, and a third (33.1%) to pages updated in the last 30 days. By publication date the picture is older: 73.2% of dated citations go to pages first published within a year, and 18.5% to pages over two years old. The link with citation counts is weak, though: −0.157 for age, −0.187 for time since update.

A negative correlation here means younger pages get somewhat more citations. Zero would mean no link, and these values sit firmly in weak territory. The median dated page has 4 citations.

The most striking number is what happens to old pages when you look at the update date. By publication date, 18.5% of dated citations go to pages over two years old. By update date, that shrinks to 4.0%. Most old pages that still get cited have been updated since they went live.

By publication date 18.5% of dated citations go to pages over two years old; by update date only 4.0%

Figure 3. Share of dated citations by time since publication and by time since last update. Pages over two years old take 18.5% by publication date and 4.0% by update date.

Show the data

Band

By time since publication

By time since last update

0–30 days

14.9%

33.1%

31–90 days

18.9%

24.9%

91–180 days

20.5%

19.6%

181–365 days

18.7%

13.7%

1–2 years

8.3%

4.7%

Over 2 years

18.5%

4.0%

How big is the freshness effect?

Small, until a page goes a year without an update. Pages first published 0–30 days before the citation average 10.9 citations each, pages 31–90 days old 10.6, and pages 91–180 days old 10.5. The drop comes after a year: 6.6 for pages 1–2 years old and 7.1 for pages over two years old.

By update date the slope is steeper. Pages updated within 30 days average 11.3 citations; pages not updated for over two years average 4.1. On our buying-intent prompts, a page published last month and a page published three months ago are cited about equally often. A page nobody has touched for two years is a different story.

Average citations per dated page fall from 11.3 to 4.1 as time since the last update grows

Figure 4. Average citations per dated page, by time since the last update and by time since publication. Averages are citations divided by dated pages in each band.

Show the data

Band

Average citations per page: by time since publication

Average citations per page: by time since last update

0–30 days

10.9

11.3

31–90 days

10.6

10.4

91–180 days

10.5

9.7

181–365 days

9.3

7.8

1–2 years

6.6

4.9

Over 2 years

7.1

4.1

Which matters more: the publication date or the last update?

The last update, by a small margin. Time since update correlates with citations at −0.187, against −0.157 for age, across 218,795 dated pages. The older a cited page is, the more likely it shows an update: 22.8% of pages under 30 days old do, against 78.3% of pages over two years old. Most old pages that still get cited have been refreshed.

Among pages that show an update, the gap is smaller (−0.167 for time since update, −0.157 for age) but points the same way. For pages that were never updated, the two measures are the same thing, and both come out at −0.203.

52.1% of dated pages show an update after first publication. The median gap between publishing and updating is 188 days, roughly six months. For anyone deciding between a new page and a refresh of an old one, this is the result that matters: the date that counts most is the one you control without starting over.

Time since update correlates with citations at -0.187, more strongly than age at -0.157

Figure 5. Rank correlation between page age and citations, by time since publication and by time since last update. All values are negative: younger pages get more citations.

Show the data


Pages

Count

Correlation: time since publication

Correlation: time since last update

All dated pages

218,795

-0.157

-0.187

Pages that show an update

114,013

-0.157

-0.167

Pages never updated

104,782

-0.203

-0.203

Do updated pages get cited more than pages that were never updated?

Yes, at every age. In all six age bands, pages that show an update have a higher median citation count than never-updated pages of the same age. The gap widens as pages get older: among pages over two years old, updated ones average 7.9 citations against 3.9 for pages that were never updated. Updated pages are 52.1% of dated pages and take 55.9% of dated citations.

The medians are small numbers, but the direction never changes: 6 against 4 for pages up to 30 days old, 6 against 5 at 31–90 days, 5 against 4 at 91–180 days and at 181–365 days, and 3 against 2 for both bands past a year. The averages tell the same story more loudly, from 13.6 against 10.0 for the youngest pages to 7.9 against 3.9 for the oldest.

"Updated" here means the page shows a later date than its publication date. We can't tell a full rewrite from a changed date stamp, so this result doesn't prove that rewriting works. What we can say is that the pages whose dates our checks flagged as suspect, mostly automatic site-wide stamps, show almost no link with citations (−0.002 by age, 0.061 by time since update). Dates that don't reflect a real change look worthless in this data.

Updated pages get more citations than never-updated pages in every age band: 7.9 against 3.9 over two years old

Figure 6. Average citations per dated page by page age, for pages that show an update and pages that don't. Over two years old, 7.9 against 3.9.

Show the data

Page age

Average: updated

Average: never updated

Median: updated

Median: never updated

Share of pages updated

0–30 days

13.6

10.0

6

4

22.8%

31–90 days

11.3

10.2

6

5

38.3%

91–180 days

11.6

9.6

5

4

45.0%

181–365 days

10.8

7.7

5

4

52.6%

1–2 years

7.9

4.6

3

2

59.4%

Over 2 years

7.9

3.9

3

2

78.3%

Which AI engine cites the freshest content?

ChatGPT. Its median cited page was updated 73 days before the citation, against 97 days for Gemini and 175 for Google AI Overviews. By publication date the gap is wider: AI Overviews' median cited page is 398 days old, ChatGPT's 160. AI Overviews sends 30.0% of its dated citations to pages over two years old, against 17.6% for ChatGPT.

Weighted by citations, the order holds: 57 days since update for ChatGPT, 79 for Gemini and 146 for AI Overviews. By update date, 93.4% of ChatGPT's dated citations fall within a year, 92.8% of Gemini's and 75.3% of AI Overviews'.

Gemini cites dated pages more often than either engine (60.9% of its citations), which fits its heavy use of listicles; most cited listicles carry a date. The correlations per engine are all weak: −0.108 by update date for ChatGPT, −0.086 for AI Overviews, and −0.143 by age for Gemini, the one engine where the publication date predicts citations better than the update date. Each engine's figure is weaker than the overall one because, in the overall figure, a page's citations from all three engines add up.

For AI Overviews in particular, an older page that's still strong keeps earning citations. Our guide on ranking in AI Overviews covers what else it rewards.

Median days since update: ChatGPT 73, Gemini 97, AI Overviews 175

Figure 7. Median days since the cited page was last updated, and since it was first published, per engine. ChatGPT 73 and 160 days; AI Overviews 175 and 398.

Show the data

Engine

Days since last update (median per page)

Days since publication (median per page)

Days since last update (weighted by citations)

Days since publication (weighted)

ChatGPT

73

160

57

135

Gemini

97

216

79

183

AI Overviews

175

398

146

340

How much AI-cited content has no date?

About a third. Pages that show no publication date anywhere our crawler looked take 32.1% of all citations. That is normal for homepages (70.6% of their citations go to undated pages) and landing pages (63.7%), but it also covers 52.0% of citations to blog and news articles. Listicles are mostly dated: only 16.6% of their citations go to undated pages.

Among the pages we could fetch, undated pages take 36.5% of citations, and articles alone account for 41.9% of all undated citations. "No date found" means none of the five places the crawler checked had one: no structured data, no meta tag, no microdata, no <time> tag and no date in the URL.

This result says as much about how sites mark up their pages as about what AI engines prefer. Still, cited pages with a date skew heavily toward recent updates, so a page without a machine-readable date gives up a signal the engines appear to reward. That is an inference from the pattern; the study can't test what an undated page would have earned with a date. Showing a date costs almost nothing. Google's guide to publication dates explains how it reads them, and a minimal version looks like this:

<p>Last updated: 2 October 2026</p>

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Best help desk software for small teams",
  "datePublished": "2025-03-14",
  "dateModified": "2026-10-02"
}
</script>
<p>Last updated: 2 October 2026</p>

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Best help desk software for small teams",
  "datePublished": "2025-03-14",
  "dateModified": "2026-10-02"
}
</script>
<p>Last updated: 2 October 2026</p>

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Best help desk software for small teams",
  "datePublished": "2025-03-14",
  "dateModified": "2026-10-02"
}
</script>
Share of citations going to undated pages: 70.6% for homepages, 52.0% for articles, 16.6% for listicles

Figure 8. Share of each page type's citations that go to pages with no publication date. Articles: 52.0%, against 32.1% for all page types.

Show the data

Page type

Citations

Share going to undated pages

Homepage

170,868

70.6%

Landing page

134,514

63.7%

Articles (blogs, news)

1,038,622

52.0%

Social

13,465

51.4%

Other (unclassified)

325,017

49.1%

Marketplace

32

37.5%

How-to

57,279

22.2%

Alternatives

25,554

16.9%

Listicle

1,859,054

16.6%

Comparison

189,895

16.1%

Reviews

144,766

11.9%

YouTube

42,212

6.3%

Forums

15,431

5.8%

Wikipedia

2,173

0.0%

Does freshness matter more in fast-moving industries?

Fast-moving industries cite younger content, but freshness predicts citations better in slow ones. Data, AI and web data has the youngest cited content (median 127 days old) and Healthcare providers the oldest (343 days). Yet the link between age and citations is strongest in Legal and professional services (−0.206) and Healthcare providers (−0.192), and flat in Fashion, beauty and consumer goods (0.003).

We grouped four industries as fast-moving (Data, AI and web data; Software and developer tools; Marketing and advertising; Banking and fintech) and three as slow (Industrial, aerospace and defence; Legal and professional services; Healthcare providers). Pooled, the fast group's cited pages are younger (median 155 days against 233), but its correlation is weaker (−0.150 against −0.202). We expected freshness to matter more where things change quickly; the data says that's where content is younger, not where freshness separates winners from losers.

By update date, all 23 industries show the same direction (−0.109 to −0.209). In consumer industries such as retail, electronics, consumer finance and fashion, the publication date barely matters (−0.063 to 0.003) while the update date still does (−0.109 to −0.188). Weighted by citations, Consumer finance and insurance has the oldest cited content of any industry (301 days).

Scatter of 23 industries: fast-moving industries cite younger content, but the age-citation link is stronger in Legal and Healthcare providers

Figure 9. Each industry's median cited-page age against its age–citation correlation. Fast-moving industries sit to the left; the strongest links are in slow industries.

Show the data

Industry

Group

Median age of cited pages (days)

Correlation: age vs citations

Data, AI and web data

Fast-moving

127

-0.126

Sales and customer experience

Other

133

-0.123

Marketing and advertising

Fast-moving

153

-0.148

Software and developer tools

Fast-moving

163

-0.165

HR and workplace

Other

167

-0.172

Banking and fintech

Fast-moving

180

-0.137

Gaming and gambling

Other

183

-0.081

Logistics and e-commerce operations

Other

184

-0.154

Cybersecurity, identity and compliance

Other

190

-0.15

Legal and professional services

Slow-moving

194

-0.206

Media and publishing

Other

195

-0.191

Electronics, hardware and telecom

Other

198

-0.059

Travel and leisure

Other

238

-0.146

Energy and automotive

Other

240

-0.123

Consumer retail

Other

242

-0.063

Real estate and home services

Other

248

-0.17

Industrial, aerospace and defence

Slow-moving

252

-0.183

Fashion, beauty and consumer goods

Other

264

0.003

Consumer finance and insurance

Other

284

-0.032

Education and public sector

Other

295

-0.189

Food, drink and agriculture

Other

296

-0.135

Health and wellness products

Other

317

-0.149

Healthcare providers

Slow-moving

343

-0.192

Which page types are the freshest?

Alternatives pages: the median cited one is 83 days old and was updated 38 days before the citation. Listicles come fourth by time since update (74 days), behind Wikipedia (31 days, where any revision counts as an update), alternatives pages (38) and homepages (68). Forum threads are the oldest cited content that isn't reference material, at a median of 559 days.

We expected listicles to be the freshest type, on the logic that "best X" roundups get refreshed every year. They are fresher than articles (232 days old), reviews (202) and forum threads (559), with a median age of 156 days, but they aren't at the top. Homepages show how much the update date changes the picture: they are very old by publication date (1,113 days) and very fresh by update date (68 days), because 89.4% of the dated homepages show an update.

Median days since update by page type: alternatives pages 38, listicles 74, forum threads 555

Figure 10. Median days since the cited page was last updated, by page type. Wikipedia counts any revision as an update; Reddit and YouTube show no edit dates.

Show the data

Page type

Days since last update (median)

Days since publication (median)

Dated pages

Wikipedia

31

6954

341

Alternatives

38

83

2,251

Homepage

68

1113

3,967

Listicle

74

156

99,045

Comparison

75

133

12,891

How-to

85

185

4,435

Reviews

90

202

5,371

Articles (blogs, news)

104

232

49,073

Other (unclassified)

163

534

22,849

Social

168

310

667

Landing page

190

543

3,575

Marketplace

300

1336

5

YouTube

337

337

9,681

Forums

555

559

4,644

What gets cited when content is over two years old?

Mostly listicles and articles. Of citations to pages first published over two years ago, listicles take 50.8% and articles 23.6%; every other page type together takes 25.6%. We expected old content to be cited mainly as reference pages such as brand pages, documentation and Wikipedia. It isn't: old content gets cited in the same formats as new content, and 78.3% of those old pages show an update.

Of citations to pages over two years old, 50.8% go to listicles and 23.6% to articles

Figure 11. Page types of citations to pages first published over two years ago. Listicles and articles take 74.4% between them.

Show the data

Page type

Citations to pages over two years old

Share of that band

Listicle

186,615

50.8%

Articles (blogs, news)

86,905

23.6%

Other (unclassified)

22,743

6.2%

Reviews

15,848

4.3%

Homepage

15,460

4.2%

Landing page

10,136

2.8%

Comparison

9,705

2.6%

YouTube

7,188

2.0%

How-to

5,774

1.6%

Forums

3,454

0.9%

Alternatives

2,031

0.6%

Wikipedia

1,542

0.4%

Social

270

0.1%

Marketplace

5

0.0%

Which of our hypotheses held up?

We wrote seven hypotheses and their pass tests before pulling any data. Four held and three didn't. The first one is an easy test to pass, since any later date counts as an update, and we say so rather than lean on it.

Hypothesis

Verdict

Deciding numbers

Most dated citations go to pages updated in the last 12 months

Confirmed

91.3% (test: over 50%)

The last-updated date matters more than the publication date

Confirmed

−0.187 against −0.157

Updated pages outperform never-updated pages of the same age

Confirmed

Higher median citations in all six age bands

ChatGPT cites fresher content than Gemini and AI Overviews

Confirmed

Median days since update: 73, 97 and 175

Freshness matters more in fast-moving industries

Rejected

Fast industries cite younger content (155 against 233 days) but correlate more weakly (−0.150 against −0.202)

Listicles are the freshest page type

Rejected

Listicles rank 4th of 14 (74 days)

Content over two years old is cited mainly as reference pages

Rejected

Listicles (50.8%) and articles (23.6%) dominate that band

What does this mean for your content?

Refresh the pages that already get cited, and show the new date. Updated pages out-cite never-updated ones at every age, and the update date predicts citations better than the publication date. By hand, that means listing your cited URLs, checking each one's visible date and its dateModified, and queueing the oldest for a real update. In AIclicks, the dashboard's Brand Pages list shows which of your pages AI engines cite, and the Sources tab filters cited URLs by engine and date range, which gives you the refresh queue. Writer 2.0 helps with the rewrites; its editing guidance names freshness as one of the three things to check.

Treat freshness as a tiebreaker. Correlations of −0.157 to −0.187 are weak. A fresh page with nothing else going for it won't be cited because it's fresh, and among comparable pages the fresher one tends to win. Don't re-date pages without real changes: the dates our checks flagged as suspect show almost no link with citations. The other factors that drive ChatGPT citations still do most of the work.

Show a date at all. A visible "Last updated" line plus datePublished and dateModified in structured data, as in the example above. This is an inference from the data rather than a tested result, but it costs almost nothing, and our guide to optimizing your website for ChatGPT covers the rest of the page-level basics.

Match the engine. ChatGPT rewards recency most. AI Overviews keeps citing older pages, with 30.0% of its dated citations going to pages over two years old, so a strong evergreen page still earns there. Track them separately with the ChatGPT tracker and the AI Overviews tracker.

Know your industry's bar. In data, AI, sales and marketing, the median cited page is 127–153 days old, about four to five months. In healthcare, legal and industrial categories, cited content is older, but the fresher pages win more clearly there than anywhere else.

Keep the URL and update the page. Three quarters of citations to pages over two years old go to listicles and articles, and 78.3% of those old pages show an update. An ageing roundup that gets maintained keeps being cited; there's no need to republish it under a new URL. For the wider playbook, see how to get cited by AI.

Methodology and limits

Item

Value

Citation data

535,560 answers from ChatGPT, Gemini and Google AI Overviews to about 25,000 buying-intent prompts across 833 product categories, 20–29 October 2026

Citations

4,018,882, capped at the number of answers each URL could have appeared in

Cited pages

527,656 URLs; 527,629 crawled and 27 not crawled

Crawl

30 October–4 November 2026; 84.1% of URLs fetched, 45.7% of URLs with a date found

Not fetched

Blocked by bot protection 9.6% of URLs, errors 4.2%, PDFs and images 1.6%, timeouts 0.5%

Suspect dates

13,736 pages flagged and kept out of every age figure

Crawl problems and fixes. A local DNS outage during the crawl logged reachable sites as missing, so we added a guard that waits out lost connections and re-crawled all 44,355 affected URLs. 2,329 URLs refused every request with a rate-limit error and stay counted as blocked.

Definitions. Age is the citation date minus the publication date; time since update is the citation date minus the last-updated date. A page counts as updated when its update date is later than its publication date. Correlations are Spearman rank correlations between age (or time since update) and citations per page, where a negative value means younger pages get more citations.

Limits. The dates are the ones publishers show: an "update" can be a real rewrite or a changed stamp, and only obvious automatic stamps are flagged. Freshness and citations move together, but the study doesn't show that updating a page causes more citations; well-run sites may simply update more and get cited more for other reasons. Every age figure describes the dated half of citations (49.4%); undated pages take 32.1%, unreadable pages 12.1%, and flagged or "date after citation" pages the rest. The pages we couldn't read aren't random: 33.1% of social, 33.0% of unclassified, 23.6% of review and 16.2% of forum citations weren't fetched, against 8.1% for listicles and 11.5% for articles. Reddit and YouTube have no edit dates, and Wikipedia's update date is its latest revision. The data holds only cited pages, so it shows how citations spread across content age, not the chance that a page of a given age gets cited. It covers one week of citations, crawled 1–15 days later; pages that showed a date later than their first citation (2.8% of citations) are reported apart because their date at the time of citation is unknown. All prompts were buying-intent questions. AIclicks sometimes counts a URL more than once per answer, so we capped each URL's citations at its prompts multiplied by its run days, which removed 6.9% of raw citations.

Cite this study. AIclicks (2026). Does Content Freshness Matter in AI Search? What 4 Million Citations Show. 527,656 cited pages crawled; citations from 535,560 AI answers, 20–29 October 2026. aiclicks.io/blog/content-freshness-ai-search-study

Frequently asked questions

Does ChatGPT prefer fresh content?

More than the other engines we tracked: its median cited page was updated 73 days before the citation, and 36.1% of its dated citations go to pages updated within 30 days. The ChatGPT tracker shows which of your pages it cites.

How often should I update content for AI search?

The data can't set a schedule, but 91.3% of dated citations go to pages updated within a year, and in data, AI, sales and marketing the median cited page is about four to five months old. More in how to get cited by AI.

Does changing the date on a page help it get cited?

Not on its own. Dates our checks flagged as suspect, mostly automatic site-wide stamps, show almost no link with citations; pages showing a genuine later update are the ones cited more. More in ChatGPT ranking factors.

Do Google AI Overviews cite old content?

Yes, more than ChatGPT or Gemini: 30.0% of its dated citations go to pages over two years old, and its median cited page is 398 days old. More in how to rank in AI Overviews.

Is content freshness a ranking factor in AI search?

In our data it's a weak signal, with correlations of −0.157 to −0.187: it helps among comparable pages rather than deciding which page gets cited. Our AI SEO statistics put it next to the other signals.

How many AI-cited pages show no date?

38.4% of all cited URLs showed no publication date anywhere our crawler looked, and they take 32.1% of all citations. More in our guide to optimizing your website for ChatGPT.

Should I republish old content under a new URL?

There's no need in this data: 78.3% of cited pages over two years old show an update and keep getting cited at their original URL. The first study in this series shows which page types those citations go to.

What to do next

Pull the list of your pages that AI engines already cite, sort it by last-updated date, and refresh the oldest ones that still matter to you. Make the update real, put the new date where both readers and crawlers can see it, and keep the URL. AIclicks shows which of your pages each engine cites and when, so you can see whether the refresh moved anything. See how AIclicks tracks your prompts.

This post is part of a three-part study built on the same data:

SEO & GEO Specialist @ aiclicks.io

SEO & GEO Specialist @ aiclicks.io

Maksim Baradulin is an SEO and GEO specialist with more than 1.5 years of experience, working with some of Lithuania’s biggest companies. At AIclicks, Maksim takes on high-priority GEO work for the platform’s largest clients. Maksim specializes in agentic SEO and GEO, and builds automations that handle repeat search and AI visibility work. The goal is brands that rank in Google and get cited in answers from ChatGPT and other LLMs.

Maksim Baradulin is an SEO and GEO specialist with more than 1.5 years of experience, working with some of Lithuania’s biggest companies. At AIclicks, Maksim takes on high-priority GEO work for the platform’s largest clients. Maksim specializes in agentic SEO and GEO, and builds automations that handle repeat search and AI visibility work. The goal is brands that rank in Google and get cited in answers from ChatGPT and other LLMs.

¿Te gustó lo que leíste? Obtén tu guía definitiva de SEO con IA aquí:

¿Te gustó lo que leíste? Obtén tu guía definitiva de SEO con IA aquí:

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

¿Te queda alguna duda?
Reserva una llamada aquí: