This article originally shared an early experiment: I asked ChatGPT broad questions about website design and published its output. The AI authorship was disclosed at the time. This edited version keeps that context while removing repetition and advice that should not be treated as verified guidance.
What the experiment suggested
The generated article recommended clear navigation, responsive layouts, consistent branding, loading performance and visible calls to action. Those are reasonable topics for a design discussion, but listing them does not establish that a particular website works well. The original output also described the model as having “opinions”. In this context, those were generated responses to a prompt, not observations from testing the website with its customers.
Where general advice falls short
A recommendation to simplify navigation does not tell you which labels confuse your visitors. A claim that a site should load quickly does not identify the resource causing a delay. A useful review needs the actual content, interface and task, with evidence appropriate to the question. Generated text can suggest questions without supplying the answers.
Check before applying a suggestion
Ask whether the advice fits your audience and whether it can be tested. Verify technical claims against relevant documentation. Do not use generated statistics, quotations or client results without evidence. Consider the information you share with a tool and the terms under which it is processed. A design question does not require uploading confidential customer records.
Keep responsibility with the reviewer
The useful lesson from this early experiment is to distinguish plausible language from an assessed recommendation. AI assistance should not hide where an idea came from or replace accountability for published claims. For a practical review of your own site, explore my website design work or contact me.
