GEO and Recency: Why AI Favors Fresh Content and How to Keep Your Evergreens Visible

AI applies a brutal time filter: why your classic content disappears

Unlike traditional SEO search engines, where freshness is one signal among many, generative AI (ChatGPT, Claude, Perplexity, Gemini) seems to apply an aggressive time filter. A fundamentally accurate article can disappear from AI responses within weeks or months if it shows no sign of being updated. Your guide on digital marketing basics, published two years ago, will no longer be cited by AI engines, even if it remains technically relevant, because these systems give disproportionate weight to recency as an indicator of reliability and contextual relevance.

This behavior is not a bug: generative AI was trained on corpora where content freshness strongly correlates with accuracy (especially for facts that change quickly: prices, technology, events). The systems therefore learned to systematically favor recent content, regardless of its actual quality.

How AI actually measures freshness

The publication date: the most visible signal, but not the only one

The publication date (or "publish date" in schema tags) is the most direct temporal signal. AI detects it through HTML metadata (the <meta property="article:published_time"> tag) and schema.org markup. However, this signal alone is not enough to keep your content visible for long.

The problem: an old publication date, combined with no detectable activity afterward, sends the message "this content has been abandoned." AI interprets the lack of maintenance as a sign of obsolescence, even if only implicitly.

The last modified date: the decisive signal but rarely used

The <meta property="article:modified_time"> tag is far more decisive than the publication date for AI. A detected modification actively signals that the creator is monitoring the content and updating it. This is the signal AI prioritizes.

However, most sites never fill in this tag correctly. As a result, AI sees no evidence of maintenance, and the content is classified as outdated even if it was slightly modified. This is an invisible and very underused lever.

Recent citations: when other content references you

AI also observes recent citations. If a new high-quality article references your old content (with a link or a citation), it signals that your content remains relevant in the current conversation. This is an indirect but powerful signal: you are part of the topic's ecosystem today, not just an archaeological remnant of the past.

You can generate this signal by creating new content that cites your evergreen resources, or by encouraging other creators to do so.

Topic evolution: contextualizing recency

AI also measures whether a topic itself has evolved. If your article covers a topic that changes (e.g., "How to use ChatGPT"), its freshness is a criterion. If your article covers a stable topic (e.g., "The fundamentals of English grammar"), age matters less.

The challenge: AI is not perfect at determining whether a topic has actually evolved or not. It overcorrects: even stable topics are treated as if they could change, hence the brutal time filter.

Why AI favors recency (and it is rational, up to a point)

The risk of misinformation: freshness as a proxy for reliability

Generative AI hallucinates. Faced with a choice between two seemingly equivalent pieces of content, it relies on reliability signals. Freshness is a proxy: if the content was published or updated recently, it is likely still true. An article on fintech industry regulations, updated last week, is more reliable than an identical article published in 2021.

This heuristic works well for topics that change quickly. It works poorly for evergreens. AI does not tell the difference.

User trust: showing that it is "still active"

When a user asks an AI "how do I optimize my SEO content?" and the AI cites an article from 2018, the user may legitimately doubt: "That's old, the rules have changed." Even if the article remains 80% relevant, the doubt destroys trust.

AI has learned (implicitly) that serving recent content improves user satisfaction. So it optimizes for that.

The limitation of training data: no "living context"

LLMs are trained on fixed data (a cutoff date). They do not have real-time access to the web. To compensate, they have learned heuristics: "content published/modified recently" = good signal. These heuristics are shortcuts, useful but blunt.

How to keep your evergreens visible without rewriting them entirely

1. Micro-updates: adding discreet freshness signals

You do not need to rewrite your article. Minor but detectable modifications are enough.

Concrete strategies:

• Add an "Update 2024" section at the beginning of the article with 2-3 contextual sentences (e.g., "The methods remain valid in 2024, although the ecosystem has evolved on these points...").
• Replace a broken link with an up-to-date source.
• Add a current example or a recent use case without modifying the body of the text.
• Update a date, a figure, or a technology mentioned as "today's."

Crucial point: every modification must be accompanied by a visible update of the <meta property="article:modified_time"> tag. Without that, the AI will see no difference.

2. Fix the modified_time tag: the invisible lever

Most CMS systems (WordPress, etc.) do not automatically refresh this tag with every modification. Check that:

• Your modified_time tag updates every time you edit the article.
• Your XML sitemap includes the <lastmod> tag with an up-to-date date.
• Your schema.org markup (article.schema) includes an up-to-date dateModified.

This is technically simple, but often forgotten. Fixing it can revive an article's AI visibility without any rewriting.

3. Create new content that cites your evergreens

Create complementary articles that reference your classic content. For example:

• An article "2024 SEO Trends" that cites your SEO basics guide as a foundation.
• A news update that links to your evergreen resource for deeper reading.
• A "before/after" piece that shows how your old method compares to new practices.

Each recent citation reinforces the signal: your evergreen content remains part of the conversation in 2024.

4. Strategic refresh: when to rewrite lightly

Every 6-12 months, plan a light refresh of your 10-20 most important articles. Do not wait for visibility to collapse.

The process:

1. Read your article with a critical eye: is it still accurate?
2. Update the examples or studies cited to make the context current.
3. Add 2-3 sentences that contextualize the content for the current year.
4. Immediately fix the modified_time tag.
5. Share the update on your channels (this signal is also detected by AI).

This operation takes 30-45 min per article, but can double the AI lifespan of your content.

5. Structure content for AI citability

AI cites well-structured content more easily. This indirectly amplifies your visibility.

• Use clear Markdown headings (h2, h3) that ask questions ("Why...", "How...").
• Start each section with a standalone sentence that summarizes the answer (AI-friendly).
• Break into short paragraphs (3-4 sentences max).
• Avoid lists of 15+ items (AI has trouble citing them partially).

Well-structured content, even old, will be cited more often than dense, unstructured content.

The maintenance calendar: staying visible without overloading yourself

You do not need to update all your articles every month. A pragmatic calendar:

• Pillar articles (top 10-15): light refresh every 6-9 months. Mini-update (1-2 sentences) every 3 months.
• Important articles (top 50): annual audit, mini-update if necessary.
• Secondary content: audit every 18-24 months, rewrite only if there are signs of obsolescence (massive broken links, replaced technology).

Common mistakes to avoid

Do not confuse "recency" with "content volume"

Creating entirely new articles every week is not necessary. AI prefers an evergreen article regularly updated over 20 abandoned drafts. Quality + maintenance >> quantity + abandonment.

Forgetting the modified_time tag

The technical modification must accompany the actual modification. Without a technical signal, the AI will not see it.

Updating for no reason

If your article on the "5 benefits of gray matter" has not changed, do not touch it just for the sake of it. Too frequent or insignificant updates are counterproductive (noise).

Neglecting structure in favor of content

An article with perfect content but poor structure (huge paragraphs, no clear headings) will be cited less often by AI. Invest in structure.

Summary: why evergreens should not be abandoned

The classic trap: create an excellent guide, then abandon it. A few months later, an AI cites a weaker but more recent competing article. You conclude that evergreen content is dead.

The reality: your evergreen content can remain viral with AI for years, provided you maintain two simple signals: (1) visible proof of maintenance (modified_time tag, micro-updates), and (2) integration into the current conversation (recent citations, new content that references you).

AI looks for fresh content, yes. But it also looks for maintained content. Old but clearly maintained content will beat new and already abandoned content.

Key takeaways

AI detects freshness through four channels: publication date, last modified date (rarely used, highly impactful), recent citations, and topic evolution context. It favors recency because it correlates with reliability and user satisfaction. To keep your evergreens visible, you do not need to rewrite them: semi-annual micro-updates, a correct modified_time tag, and new content that references you are enough. Invest 30-45 minutes every 6-9 months rather than rewriting entirely, and your content will remain AI-visible in the long term.