Testing Different AI Disclosure UX Patterns, What Users Actually Preferred

A client needed to disclose that product descriptions on their site were AI-generated, both for transparency and increasingly for compliance reasons. We tested three disclosure patterns against each other, expecting the most prominent one to win on trust. It didn't.

The three patterns tested

• A bold banner at the top of the page: 'This content was generated with AI assistance.'
• A small, consistent inline badge next to each AI-generated block, unobtrusive but always present.
• A specific one-line explanation of what the AI did and didn't do: 'This description was drafted by AI from the manufacturer's spec sheet and reviewed by our team.'

Users didn't want to know that AI was involved, they wanted to know specifically what it did — and a vague disclosure read as less trustworthy than no disclosure at all in our panel.

What the testing showed

The bold banner actually scored lowest on subsequent trust ratings — several participants described it as feeling like a legal disclaimer rather than useful information, which made them more suspicious of the content, not less. The specific one-liner scored highest, because it answered the actual question users had, which wasn't 'was AI involved' but 'how much should I trust this and why.'

The inline badge alone, without the explanatory text, landed in the middle — present enough to satisfy transparency requirements but vague enough that it didn't meaningfully move trust in either direction. We shipped the specific one-liner pattern, scoped down to a short badge-plus-tooltip to avoid cluttering every product card, and post-launch surveys showed trust in the content held steady compared to a pre-AI baseline, which was the actual goal — disclosure that didn't cost trust it didn't need to cost.