# Google research: entity order limits LLM fact recall

Published: 2026-08-17T11:17:13.208Z · Source: Search Engine Journal (https://www.searchenginejournal.com/google-subject-object-entity-order-affects-ai-answers/586089/)
Source date: 2026-08-17
Entities: gemini, google

A new Google research paper finds that encoding isn't the bottleneck for LLM fact recall — frontier models including Gemini-3-Pro and GPT-5 encode 95-98% of facts but fail to directly recall 26-34% of them, per Search Engine Journal. The paper ties part of the gap to subject/object entity order: models trained on a fact phrased one way struggle to answer when a question reverses that order, and recall failures account for more than 70% of GPT-5.2's errors. Extended "thinking" recovered 40-65% of the otherwise-unrecalled facts, per the paper.

Why it matters: If phrasing shapes which subject/object order a fact gets encoded in, how a fact is worded — not just whether it's present — may affect whether an AI answer engine can retrieve it, though the researchers stop short of confirming this as a content lever.

## What this answers

**Why do AI chatbots sometimes fail to answer questions even when they "know" the fact?**

A Google research paper finds frontier LLMs encode 95-98% of facts but fail to directly recall 26-34% of them — the bottleneck is retrieval, not missing knowledge, and reversing a fact's usual subject/object order makes recall harder.

**Does asking an AI model to think longer improve its fact recall?**

Yes — extended "thinking" recovered 40-65% of facts models could not otherwise recall directly, per the Google research covered by Search Engine Journal.


Canonical: https://anythingengineoptimization.com/item/2026-08-17-google-research-entity-order-limits-llm-fact-recall/
From Anything Engine Optimization (AEO Wire) — https://anythingengineoptimization.com/ · Standards: https://anythingengineoptimization.com/standards/