Does content freshness matter to AI search engines even when the content itself does not change?
We tested the smallest possible version of that question.
On August 4, 2026, we updated the last published date on 10 existing OtterlyAI articles. We did not change the content itself.
Over the following 30 days, citations to those pages increased 31.5%. Two unchanged control articles moved in the opposite direction, declining 10.6%, from 358 to 320 citations.
Key findings
- 31.5% more citations: the 10 updated articles increased from 4,800 to 6,312 citations in the 30 days after the date change. The control pages declined 10.6% over the same period.
- 41 more prompts cited the updated articles: The number of distinct prompts citing at least one article rose from 205 (64%) to 246 (76%), an increase of 12 percentage points.
- 7 of 10 updated pages gained citations: one remained unchanged and two declined. Excluding the highest-volume article, the remaining nine still recorded a 30.4% increase.
- ChatGPT and Claude increased sharply: ChatGPT citations rose 692%, while Claude increased 87%.
What we tested
Publication dates are an obvious freshness signal for readers and search engines. But does changing that signal alone make any difference to how often an article is cited by AI search engines?
To isolate it, we selected 10 existing OtterlyAI articles with older visible update dates. Several still displayed dates from 2024 or 2025, while others had been published earlier in 2026. On August 4, 2026, we changed only the visible “Last updated” date on all 10 pages. The underlying content remained unchanged.
We compared citation performance for the 30 days immediately before and after the change:
The comparison asks a deliberately narrow question: what happened to AI citations when the publication date changed but the article did not?
The experiment covered a US brand report monitoring 320 prompts across seven AI engines, representing an estimated 67,200 AI answers over the initial 30-day period.
Citations increased 31.5% across the updated articles
The clearest result was in total citation volume.
The 10 updated articles received 4,800 citations in the 30 days before the experiment and 6,312 in the 30 days after it. That is a 31.5% increase.
The two unchanged control articles moved in the other direction, from 358 to 320 citations, a 10.6% decline.
For a change that required no rewrite of the underlying content, that is a meaningful difference.
It was also reasonably broad across the sample. Seven of the 10 updated articles gained citations, one remained flat and two declined.
And the result was not simply caused by the highest-volume article in the experiment.
That page increased from 3,624 to 4,778 citations. But when we remove it entirely, the other nine updated pages still increased from 1,176 to 1,534 citations, a 30.4% increase.
In other words, the overall uplift remained almost identical without the experiment’s largest page.
The increase also extended across a broader range of tracked prompts. Before the update, 205 of the 320 monitored prompts (64%) cited at least one of the ten articles. After the update, that number rose to 246 prompts (76%), an increase of 41 prompts and 12 percentage points.
To make the treated and control groups comparable despite their different sizes, the chart below shows average daily AI citations and Google organic clicks per article rather than group totals.
On a daily per-article basis, the treated group showed an increase in both AI citations and Google organic clicks after the update, while the control group declined on both metrics.
Citation gains varied significantly across AI engines
ChatGPT recorded the largest percentage increase, rising 692%, from 37 to 293 citations. However, it started from a relatively low baseline, and the increase represents 256 additional citations.
Claude delivered the largest absolute gain, adding 983 citations, from 1,135 to 2,118 (+87%).
Perplexity also saw substantial growth, adding 674 citations (+217%). It was the only engine where no updated article experienced a decline: eight of the ten pages gained citations, while two remained unchanged.
Google AI Overviews recorded a more modest increase of 177 citations (+15%). In contrast, Google AI Mode and Gemini declined by 20% and 37%, respectively.
Despite these differences, total citations across all seven engines increased 31.5%. This highlights the importance of looking at each engine individually, as aggregate results can hide significant differences in performance.
Google clicks moved in the same direction
We also checked Google Search Console as a secondary signal.
Clicks to the 10 updated pages increased from 80 to 132, a 65% increase. The control pages declined -45% over the same period.
The absolute traffic numbers are small, so we would not make organic clicks the main finding from this experiment. But the direction was consistent with the citation result: the updated group increased while the control group declined.
Why freshness matter in AI search
AI search engines increasingly supplement what their models already know with live web searches when a question calls for current information. Some of those searches are explicitly time-sensitive. That means a user asking for “the best AI visibility tools” might trigger an underlying search for “best AI visibility tools 2026”, even if the user never mentioned a year.
This offers one possible explanation for our results. Several pages in our experiment still carried visible dates from 2024 or 2025 before we updated them. A more recent date may have made those pages a better match for recency-sensitive searches and therefore more likely to enter the pool of sources an AI engine could cite.
We did not capture the fan-out queries used for our measured citations, so we cannot say this mechanism caused the increase. But the pattern is consistent with what we know about how AI search systems retrieve current information.
What this means for GEO
The takeaway is not to change publication dates on stale content without reviewing it. Our experiment changed only the date because we wanted to isolate freshness as a variable.
In practice, the better approach is to treat freshness as part of content maintenance. If an older article is still strategically important, review it, update anything that is no longer accurate, and make its visible publication or update date reflect that work.
This may matter particularly for topics that change quickly. AI search systems frequently go to the live web for current information, and their underlying searches can explicitly look for the current year. A page that still signals “2024” may therefore be competing for retrieval against sources that signal “2026”.
Our experiment adds another piece of evidence to that picture. After updating the visible date across 10 older articles, citations increased 31.5%, with the strongest gains on ChatGPT, Perplexity and Claude.
For GEO teams, the practical lesson is simple: do not treat evergreen content as “publish once and forget.” Keep important pages genuinely current, and make that freshness visible.
Method and limitations
We selected 10 existing articles on otterly.ai and changed their last published date to August 4, 2026. No other content changes were made to those articles as part of the experiment.
Citation counts were compared across two equal 30-day periods: July 5 to August 3 before the change, and August 5 to September 3 after it. August 4, the day the dates were changed, was excluded.
Citation data covers ChatGPT, Claude, Perplexity, Google AI Overviews, Google AI Mode, Gemini and Copilot as measured in OtterlyAI.
The control group contains two pages that were left unchanged during the measurement period.
The experiment shows what happened after one tightly scoped change across these pages. We will continue testing freshness across larger page sets to see how consistently the pattern holds.
Got a GEO experiment idea? Let’s collaborate
At OtterlyAI, we don’t just write about AI Search. We test it. Together with the community, we maintain a public GEO Experimentation Tracker showing what we are testing and what appears to move AI Search visibility.
If you are running your own GEO experiment, email thaylise.nakamoto@otterly.ai. We may feature your experiment in the tracker, with full attribution.
👉 Want to measure your AI Visibility? Start with OtterlyAI and our GEO Guide.
👉 Learn how to connect your Claude or ChatGPT to OtterlyAI’s MCP
