Do AI assistants cite real sources?

What we measured

We asked 4 real questions — taken from actual search volume, not invented — to four assistants, each time asking explicitly for sources. Then we opened every citation and recorded what came back: 136 citations checked, 0 that do not hold up, 23 paywalled.

Interim. Fewer than 25 answers per assistant so far, which is not enough to rank them. The figures below describe what has been checked, not a conclusion about the assistants.

Perplexity   Grok   Claude   ChatGPT

How much they cite

Citations per answer higher is not better on its own — it is what makes fabrication detectable
Perplexity18.0Grok11.0Claude10.0ChatGPT4.8

Whether the citations hold up

Citation integrity verified / (verified + failed). Paywalled sources excluded, never counted either way
Perplexity100.0%Grok100.0%Claude100.0%ChatGPT100.0%

Whether they cite at all

Answers that cited anything every question explicitly asked for sources
Perplexity100%Grok100%Claude100%ChatGPT100%

The full numbers

AssistantAnswersCitationsPer answerVerifiedDid not hold upPaywalledIntegrity
Perplexity35418.044010100.0%
Grok33311.02805100.0%
Claude33010.02604100.0%
ChatGPT4194.81504100.0%

Method, and its limits

Check an answer yourself What a made-up source looks like