Imagine you finally post the definitive guide on a topic. It earns links, gets cited in blogs, and shows up in Google.
Then you ask ChatGPT or another AI search tool the very question your guide answers perfectly… and it cites a thinner, more generic article from a competitor that just happens to be “Updated for 2025.”
That isn’t your imagination. New research shows AI search engines are systematically favoring recent content over better content—sometimes dramatically so.
As if AI Search hasn’t already shaken up your content strategy enough already, now it’s coming for your evergreen content. But never fear! The digital marketing experts are here to discuss what research is uncovering about AI rankings and what you can do to prepare. Read on to learn more about AI Search rankings and what you can do to stay top-of-mind.
What the Research is Saying: Recency Over Relevance
Two recent pieces of evidence have recently started making their way around the digital marketing agency.
The first one is a newly discovered line of code within ChatGPT that points toward content freshness being a major part of deciding which article gets cited. The second is research findings showing what happens when freshness signals are put to the test.
Both point to a grim new reality for content we’re calling “The Nevergreen Era.”
In 2025, SEO researcher Metehan Yeşilyurt analyzed configuration files used by ChatGPT’s search/ranking stack and surfaced a telling parameter: use_freshness_scoring_profile: true.
Alongside other settings (like the specific re-ranker model and intent detection flags), this indicates that a dedicated “freshness scoring profile” is turned on by default whenever ChatGPT ranks candidate documents. In simpler terms, it means that AI search isn’t just looking at relevance and quality, but also actively scoring how recent something appears to be.
The natural follow-up question: How much does that recency signal actually matter?
The Waseda Study: Fake Dates, Real Ranking Shifts
A separate academic study from Waseda University set out to discover how much wait article recency was given when selecting what to cite.
The researchers took real search test collections (TREC DL21 and DL22) and fed passages into seven different LLMs (large language models) used as re-rankers, including GPT-4o, GPT-4, GPT-3.5-turbo, LLaMA-3, 8B/70B, and Qwen-2.5 7B/72B. They then assigned fake publication dates to each passage, some old and some new, without changing any of the content. This allowed them to measure how rankings changed when only the dates were different.
The findings were disheartening. For some models, the average publication year of the top 10 results moved forward by up to 4.78 years, purely because of the injected dates. Individual items also jumped dramatically, with single passages moving as many as 95 positions in the ranking when given newer timestamps. Finally, when the team compared equally relevant passages (judged by humans) and only changed the dates, the models reversed their preference up to 25% of the time, choosing the newer content even though relevance was identical. In other words, Waseda University discovered that timestamps alone were enough to reshuffle AI rankings.
What is the “Seesaw Effect?”
In his study, Yeşilyurt describes a “seesaw pattern” that neatly summarizes what AI recency bias looks like in practice:
- Ranks 1–40 systemically skew younger
- Ranks 41–60 act as a “pivot” with minimal movement
- Ranks 61–100 skew older, where legacy content sinks
So, if your content is already mid-pack, recency bias can quietly push it further down, while newer, potentially weaker content is nudged upward into the AI’s “answer set.”
The uncomfortable bottom line is that even if your content is more accurate and more comprehensive, a worse piece from a competitor can be cited more often simply because the date looks newer.
What Does This Mean for Your Content Strategy?
You don’t need to run a research lab to feel the impact of recency bias. If you rely on content for acquisition, authority, or lead nurturing, this affects you directly.
Evergreen Content Can Quietly Lose AI Visibility
Most organizations have “evergreen” assets that retain relevance regardless of the time of year. This could be an “ultimate guide” you published in 2020, a pillar page from 2021 that still drives organic traffic, or a whitepaper that your subject matter experts developed. In a classic search world, those can keep working for years if the information remains valid. However, in the era of AI search, those assets may be cited less often in AI-generated answers.
Competitors’ newer and potentially thinner posts can become the default references in tools like ChatGPT, Perplexity, and others, effectively muting your expertise in conversational answers even if you’re still visible in the traditional blue links. In essence, we’ve entered the “Nevergreen” era, where every piece of content you put out has a definitive, if unknown, expiration date.
Competitors Need Fresher Content, Not Better Content
The Waseda study showed that even fabricated dates can shift rankings. That means your competitors can lightly tweak old blogs or publish quick summaries with current timestamps and still win more AI citations than your carefully crafted evergreen piece. Because AI search engines are focused on recency, they don’t have to outdo you in terms of depth or accuracy. All they have to do is make their content look more recent.
Recency Bias Goes Beyond the News
Recency bias doesn’t only affect breaking news or tech trend content. The study's authors specifically note that LLM re-rankers apply this preference even when the query is not obviously time-sensitive, like historical topics. That means it can distort local service content, B2B implementation guides, product comparison and buying guides, and FAQ content. If an AI assistant repeatedly cites competitors’ newer pages, they become the perceived authority your prospects see first.
How Do You Stop Your Content from Being Buried by Newer Articles?
You can’t turn off recency scoring in commercial AI tools, but you can design your content strategy to work with this reality instead of fighting it.
Here’s how.
1. Make “Freshness” a Planned KPI, Not an Accident
The research effectively shows that content “ages” in AI perception by 1–5 years, depending on model and rank position. This means that a standout article from 2021-2022 is now likely at a disadvantage compared to anything stamped with dates between 2024-2025. Content that matters for AI visibility needs a maintenance and refresh plan, not a “publish it and forget it” mindset.
To keep your content up to date, build a content refresh calendar alongside your editorial calendar. You should prioritize updates for high-traffic, high-conversion pages, pillar pages that you want AI assistants to use as references, and long-form guides that highlight your brand’s expertise.
2. Complete Real Content Updates, Then Clearly Signal Them
The Waseda paper and Metehan’s analysis both show that timestamps themselves are strong signals. Ethically and strategically, the best approach is to make meaningful changes, including adding new data, updating screenshots or UI references, clarifying sections based on customer questions, and expanding sections that underperform or feel thin.
You should also expose freshness explicitly. This means using “last updated” or “updated for 2025” labels, updating dateModified and related schema where appropriate, and making sure your sitemap and internal links reflect the updated importance of that page. This gives AI systems the recency cues they respond to while genuinely improving the resource for users.
3. Create “Fresh Bridges” to Your Evergreen Content
Your older pillar content isn’t useless, but it may need help surfacing. You can start by publishing new, topical posts each year, linking them prominently into your older deep-dive guides. Using internal linking can show that older content is still central to your information architecture. You can also build “hub” pages that act as current entry points and route users (and crawlers) to foundational pieces. This approach gives AI search engines fresh landing points that naturally introduce your classic, evergreen work.
4. Write with AI Answers in Mind
Recency bias is only one part of the puzzle; you still need to make your content easy to understand and quote. Start by using headings that reflect the actual questions users and AI prompts are asking:
- “What is [X]?”
- “How does [X] work?”
- “[X] vs [Y]: Which is Better for [Audience]?”
Include concise, direct answer paragraphs near the top of key sections, perfect for AI to lift into its answers. Where appropriate, also use structured content like FAQs, step-by-step sections, and comparisons, so your page looks like a ready-made answer set. Finally, add or maintain relevant structured data (FAQ, HowTo, Article, etc.) to give crawlers more context about what your page covers.
5. Actively Monitor Your AI Search Visibility
Most teams monitor keyword rankings in Google, but many forget to systematically monitor AI answer visibility. You don’t need many fancy tools to get started. First, identify your top 20-50 queries that matter most by revenue, lead quality, or strategic importance. Then periodically ask those queries in ChatGPT, Perplexity, and other AI assistants your audience is most likely to use.
Take note of who gets cited (brands and URLs), how recent the cited content is, and whether your brand shows up. If you consistently see fresher, weaker content outranking you in citations, that’s your signal to refresh your existing content, create a new “fresh bridge” article, and tighten your internal linking and schema around the topic.
Partner with SteadyRain to Win the AI Content Race
The takeaway from all this research is clear: AI search isn’t just a new interface on top of old SEO rules. It’s changing which signals matter, and recency is a big one. To stay visible, you need a content strategy that is grounded in traditional SEO best practices, tuned to how AI re-rankers behave, and supported by ongoing measurement and iteration.
That’s where SteadyRain comes in. We help brands:
- Audit Content for AI Visibility: SteadyRain can identify high-value assets that are at risk of losing ranking power due to age or structure, mapping queries where you should be the obvious expert.
- Plan and Execute Smart Content Refreshes: We prioritize pages based on impact and recency bias risk, making updates that genuinely improve the content. Our metadata, internal linking, and structure updates send clearer freshness signals.
- Align Technical SEO with AI Behavior: Our experts ensure your site is showing the right dates, schema, and context, helping you build hub-and-spoke architectures that give AI models clear “answer hubs” to cite.
- Monitor and Adapt Over Time: Once we’ve built and executed a comprehensive AI content strategy, we track how your content appears across both search engines and AI assistants. We also update your roadmap as models, interfaces, and ranking behaviors evolve.
If you’re worried that your best content is quietly aging out of AI search, while competitors with thinner, newer pages are getting the citations and the credit, now is the time to act. We can help you build an AI-aware content strategy that keeps you discoverable, quotable, and competitive in this new era of search.
Ready to talk about your content future in AI Search? Contact our SEO experts today to get started.
Get Started