A note before the fix. This one does not age, because it is about how you hand work over rather than which buttons you press. It works the same in ChatGPT, Claude and Gemini.
Imagine walking past a new colleague's desk on their first morning, saying “do something about the away day”, and continuing down the corridor. Whatever lands in your inbox that afternoon will be sensible, generic and useless, and you would not blame the colleague for it because you gave them nothing to work with. That corridor moment is, more or less, what most people do to ChatGPT, and the bland answer is simply the tool holding up its end of a bargain nobody meant to strike.
Meanwhile, somebody at work keeps getting genuinely good output from the same tool, so the natural conclusion is that they know a trick. What they actually know is a habit, which is close but importantly different, and learnable in about a minute.
The short answer. ChatGPT gives generic answers because a one-line question tells it nothing about you, so it produces the safest plausible response for the average person, and the average of everyone is beige. The fix is not clever prompt wording. It is a different kind of handover: briefing it the way you would brief a capable colleague on day one, with the job, the context, the audience, the format and what a good result looks like.
Why the answers come out beige
The tool's default move is the middle of the road. When it knows nothing about who is asking or why, the statistically safe answer for everyone is all it can produce, and safe-for-everyone is generic by definition.
The quality of what comes back mostly mirrors the quality of what went in. A search-engine question gets a search-engine answer, and nothing in that exchange is broken: the machine did what a total stranger with no context would do, which is roughly what your corridor colleague did with the away day. It is also why AI can sound nothing like you while still producing perfectly grammatical prose.
From prompting to delegating
The shift is small, but it changes everything downstream. A prompt is a question, while a delegation is a handover, and a good handover to a human has always contained the same handful of things: what the job actually is, the background they will need, who the result is for, and what good looks like when it arrives.
ChatGPT responds to that structure for the same reason a person does. It finally has a target other than the average. Five lines cover it, and they are the same five every time.
| The line | What it gives the tool | For example |
|---|---|---|
| The job | What you want produced, not the topic | “Draft the email”, rather than “help with the email situation” |
| The context | The two or three facts a colleague would need | “Their late feedback caused the delay, but the relationship matters more than the point” |
| The audience | Who receives this and what they care about | “A director who hates fuss” |
| The format | Length, structure and tone | “Under 120 words, warm but direct” |
| What good looks like | The finish line, which does the most work of the five | “She agrees to Wednesday without anyone apologising twice” |
Put those five together for the deadline email and what comes back is a different species of draft from the one-line version, from the same tool, on the same afternoon. The last line earns special mention because almost everyone skips it. A tool that knows what winning means can actually aim for it, and that is what converts a competent answer into a useful one.
If you find yourself repeating the same context, custom instructions can carry the permanent parts, such as your role, audience and preferred style. The brief still supplies what is particular to today's job.
The lazy-day version
On the days you cannot face writing a proper brief, one sentence buys most of the benefit: “Before you answer, ask me the three questions that would most improve your response.” It will pull the context out of you instead of guessing at it, and the guessing is where the generic comes from.
It is also strangely clarifying, because the tool's three questions are usually the three things you had not decided yet.
When beige is fine
Judgement runs in both directions, as ever. A quick definition, a list of options to react to, or a rough starting point you intend to demolish are all perfectly served by a one-line ask, and writing a full brief for them is its own kind of time-wasting.
Save the five lines for anything carrying your name, anything going to a person who matters, and any job you would properly brief a human for. Matching the handover to the stakes is the actual skill here, and it is the difference between operating a tool and receiving things from one.
Not sure where to start?
The right brief needs the right tool underneath it.
The free finder tells you which AI to open for the task in front of you, in about 60 seconds, and the full “AI, sorted.” reference lands in your inbox.
Try the free finder →One question in, one answer out.
Clair helps non-technical professionals know when to trust their AI, when to check it, and when to skip it.