Cited but Not Verified: leaderboard
Large language models (LLMs) power deep research agents that synthesize information from hundreds of web sources into cited reports, yet these citations cannot be reliably verified. Current approaches either trust models to self-cite accurately, risking bias, or employ retrieval-augmented generation (RAG) that does not validate source accessibility, relevance, or factual consistency. We introduce
Metric: Relevant Content (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 95.7 | #83 |
| 2 | GPT-5.4 | 93.7 | #79 |
| 3 | GPT-5.2 | 92.3 | #110 |
| 4 | Claude Haiku 4.5 | 91.1 | #280 |
| 5 | Claude Sonnet 4.6 | 89.8 | #90 |
| 6 | Claude Sonnet 4.5 | 88.3 | #142 |
| 7 | GPT-5 Mini | 87.4 | #186 |
| 8 | Claude Opus 4.6 | 83.9 | #63 |
| 9 | Gemini 3 Flash | 82.9 | #99 |
| 10 | Gemini 3.1 Pro (Preview) | 80.7 | #58 |
| 11 | GPT-OSS-120B | 68.7 | #341 |
| 12 | Llama 4 Maverick | 60.6 | #464 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cited-but-not-verified · How It Works · Data refreshed daily, snapshot 2026-10-20.