AI recency bias

AI recency bias · Freshness bias in AI engines

AI recency bias is the tendency of language models and AI engines to rank recently dated content higher, at equal quality. A newer date alone can push a passage up, even when older content is more complete.

Standalone definition · 37 words

Category: Retrieval & LLM · Published on November 11, 2026 · Updated on November 11, 2026 · By Aslane Samai, SEO & GEO consultant

How Metehan Yeşilyurt spotted recency bias in ChatGPT

In August 2025, Metehan Yeşilyurt found a use_freshness_scoring_profile: true setting in ChatGPT search configuration files. He concluded that ChatGPT actively favours recent content in its ranking.

What the Waseda University study shows about recency

A Waseda University team added fake publication dates to identical passages and tested seven models (GPT-4o, GPT-4, GPT-3.5, LLaMA-3, Qwen-2.5). As reported by Metehan in October 2025: the top 10 shifts up to nearly 5 years toward recent content, a passage can jump 95 positions, and about 1 relevance decision in 4 flips on the date alone.

How to account for recency bias on your site

Date content clearly, update it for real and show the update date (see content freshness). Changing a date without changing the substance is risky with Google, which asks that dates reflect real changes.

Frequently asked questions about AI recency bias

Does recency bias exist in Google too?

Google applies freshness depending on the query type (news, evolving topics). The bias measured by the Waseda study concerns language models reranking passages.

Should you republish your articles every month?

No. Updating what changed is enough. A new date without new content adds nothing and can hurt trust.

Terms related to AI recency bias

Further reading on AI recency bias

Sources for the AI recency bias definition