Neural Toons Original

AI's 'Helpful' Sources? Mostly Fluff!

08 August 2026 · The Neural Feed
AI's 'Helpful' Sources? Mostly Fluff! – Neural Toons comic

“When AI tries to be a know-it-all, it just becomes a show-it-all.”

The story behind the toon

What happened: A new study found that a specific type of AI, called GraphRAG, over-cites in its answers. It provides 11-15 sources per answer, but only a tiny fraction (0.12-0.23) are actually relevant. This makes the AI much less reliable, especially when it needs to combine information from multiple steps to answer complex questions, with performance dropping significantly.

This Neural Toon is based on the story “RAG study finds GraphRAG over-cites and fails multi-hop tasks”.

Read our summary of the story · original source

Comic transcript

Panel 1 – Ravi: It tries to be helpful... maybe too helpful.

Panel 1 – Elena: Ugh, my AI gave me 20 links!

Panel 1 – Fact: Ever ask AI for a simple answer and get a HUGE pile of links?

Panel 2 – Ravi: GraphRAG, an AI search, gives 15 sources... but only 1 is relevant!

Panel 2 – Elena: FIFTEEN? For one good answer?!

Panel 2 – Fact: New study shows one AI type gives 15 sources, but only 1 is actually useful!

Panel 3 – Ravi: Exactly. For complex questions.

Panel 3 – Elena: So I'm wasting time checking useless links?

Panel 3 – Fact: This means you waste time checking bad links and can't trust AI for complex questions.

Panel 4 – Ravi: This AI struggles connecting facts. Reliability drops 74% to 40%!

Panel 4 – Elena: So it's bad for important decisions!

Panel 4 – Fact: This problem gets much worse when AI needs to connect many different facts.

More Neural Toons · Neural Snaps · The Neural Feed