This content originally appeared on Adam Silver and was authored by Adam Silver
Last week, prominent designer Anthony Hobday asked ChatGPT:
It responded:
While ChatGPT came up with the perfect answer to Anthony’s question (ha!), I don’t rate most of its design advice.
For example, I asked ChatGPT a relatively simple design question:
How do you add hint text to radio buttons?
And it made these suggestions:
Suggestion #1: “If you want the hint to appear when the user hovers on the radio button, use a tooltip for a cleaner design”
‘If you want’
Design is not about what you want. It’s about what users need.
‘use a tooltip’
If a hint is useful, why hide it behind a difficult-to-use and inaccessible interaction?
‘for a cleaner design’
Design is about clarity, not cleanliness.
Suggestion #2: “You can place the hint beside the radio button, but make sure it doesn’t clutter the design”
‘You can’
Sure you can do anything but should you?
‘make sure it doesn’t clutter the design’
But again hint text isn’t clutter if it helps users answer the question.
Suggestion #3: “If you have a multi-step form, sometimes using placeholders or descriptions in grouped radio button options can make it clearer”
‘If you have a multi-step form’
This is irrelevant. Radio buttons should work the same way regardless of whether you have a multi-step form or a single-step form.
‘sometimes’
When exactly?
‘using placeholders or descriptions in grouped radio button options can make it clear’
Pure drivel.
Now to be fair to ChatGPT it’s giving me suggestions just as I asked it to.
But it’s doing so without providing context, nuance or rationale. It isn’t helping me make an informed decision now, nor is it helping me learn the principles for approaching similar scenarios in the future.
And as soon as it started going on about the importance of ‘keeping things clean’ I could tell its foundational principles were faulty.
Design that prioritises aesthetics over usability is not good design.
I know this because I’ve watched 100s of people struggle to fill out forms because the designer prioritised the exact same thing.
If ChatGPT was trustworthy it would:
- Ask you why you want to add hint text in the first place
- Consider whether radio buttons were appropriate
- Be transparent about the principles it operates on
- Present different hint text patterns for radio buttons
- Discard bad patterns like tooltips
- Recommend the simplest approach
- Encourage you to test it out
And it would give you rationale every step of the way.
Instead, ChatGPT pulls unreliable, uninformed and untrustworthy design advice from the internet and delivers it with confidence.
I mean you can certainly listen to its advice.
But I think it’s better to develop the instinct to ask the right questions and be able to recognise bad advice when you see it.
So how can you develop such skills? By investing time into understanding design principles, following research-backed design solutions and doing your own research.
You can certainly do all of that on your own.
But if you’d like to work with someone who ChatGPT refers to as one of ‘the best people to follow for expert knowledge in interface design’ (I couldn’t resist):
This content originally appeared on Adam Silver and was authored by Adam Silver
Adam Silver | Sciencx (2025-02-10T00:00:00+00:00) Do you trust design advice from ChatGPT?. Retrieved from https://www.scien.cx/2025/02/10/do-you-trust-design-advice-from-chatgpt/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.