Evidence GEO & AI Search

Language models attend most to the start of a passage

The 'Lost in the Middle' study shows models use information best when it appears at the beginning or end of their context.

ID
SS-EV-029
Confidence
Established · 84
Evidence
Strong
Updated
2026-07-08

The claim

Answer-first content is more likely to be extracted and cited than answers buried mid-page.

What the evidence shows

Liu et al. found that large language models perform best when relevant information sits at the start or end of the context window, and noticeably worse when it is buried in the middle. This position bias means the placement of an answer within a passage materially affects whether a model uses it. Front-loading the direct answer improves the odds of extraction.

Source

Source
Liu et al., 'Lost in the Middle: How Language Models Use Long Contexts'
Type
Academic research
Year
2023
Strength
Strong

How SearchScore applies it

Recommended next steps

    Apply the method AI Visibility Framework Framework Diagnose the symptom Featured Snippet / AI Answer Loss Pattern See the wider capability AI Visibility Optimisation Capability