# Self-verifying AI flags its own unsupported claims Thesis AI generates long literature reviews (up to ~80 pages) with **built-in citation verification** that flags claims as unsupported, partially supported, or contradictory — making the hallucination problem visible inside the output rather than leaving it to the reader. The transparency is the feature: the first AI lit-review tool to surface its own unreliability. Trade-offs: ~40-minute generation, text-heavy output, and the risk of offloading foundational research skill to the tool. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Applies to AI-assisted academic literature reviews. - Source pointer: Thesis AI demo and citation-verification pass. ## Links - [[3. Resources/Permanent Notes/The Maker Should Not Grade Its Own Homework]] — contrasts: a model surfacing unsupported claims improves transparency, but its self-check still cannot replace independent verification. - [[3. Resources/Permanent Notes/AI-Assisted PKM Replaces Synthesis Work With the Appearance of Synthesis Work]] — applies-in: citation flags expose one form of unsupported output while leaving the risk of outsourcing foundational synthesis intact. - [[3. Resources/MOCs/AI Agent Safety MOC]] — applies-in: the MOC lists this note under Verification & the maker/checker split. ## Source - [[3. Resources/Readwise/Podcasts/This AI Tool for Literature Reviews Catches Itself Lying (AI Thesis Writing)|This AI Tool for Literature Reviews Catches Itself Lying]]