Claude answers are less accurate than ChatGPT's, study finds
Profound analyzed more than 158,000 AI-generated brand claims through its FactCheck tool and found inaccuracy rates of 4.3-6.8% across engines, with Claude more than 1.3 times as likely as ChatGPT to produce an inaccurate claim, per an August 17 report based on data through August 14. Pricing and billing was the single largest driver of errors — 12% of evaluated claims but 24% of inaccurate ones, and the top error theme for 66% of brands with at least ten pricing claims checked. Claims citing competitor sites and earned media had the highest inaccuracy rates among source types.
Why it matters: A stated methodology and six-figure sample turn 'AI answers are sometimes wrong' into something brands can act on — which topics and source types to audit first, and which engine to watch more closely.
Posted to the wire August 19, 2026.