AI Guides › Getting Started
Why New Users Over-Explain (And How To Stop)
By Nigel Guy · 2 min read
New users write prompts like legal disclaimers. Three paragraphs of
background, a list of caveats, a restatement of the obvious, and then,
somewhere near the bottom, the actual question. It feels responsible — you're
being thorough, giving it everything it needs. What it actually does is bury
the one sentence that mattered under a pile of context the model didn't need
and may not even use.
The rule: say the outcome you want first, in one sentence, then add only
the context that changes the answer.
Why it happens
You're used to writing for a reader who might miss something if you don't
spell it out, or used to search engines, where more keywords used to mean
better results. Neither instinct transfers well. A model reads the whole
prompt either way — over-explaining doesn't protect you from a bad answer,
it just makes the actual ask harder to locate, for the model and for you when
you reread it later.
The mechanism: outcome, then context
- Lead with the outcome, as a single sentence. "Write a two-paragraph
summary of this for a non-technical reader" beats three sentences of
throat-clearing before the ask shows up.
- Add only the context that would change the answer. Audience, tone,
length, and format usually qualify. Your job title, the history of why
you're doing this, or a defence of why the question is reasonable, usually
don't.
- Cut any sentence that restates something the model can already see.
If you've pasted the document, you don't need to also describe what's in
it.
- Ask for the short answer first, detail second, if you're not sure how
much you need. "Give me the answer in two sentences, then expand if I ask"
gets you a usable response immediately and lets you pull more only where
you actually want it.
- Reread your prompt once before sending it and delete anything that's
there to make you feel thorough rather than to change the output.
What to skip
Skip apologising in the prompt for anything — "sorry if this is a silly
question," "I know this might be basic." It changes nothing about the answer
and costs you nothing to remove. Skip restating the full history of a
back-and-forth every message; in a single conversation, the model already has
that context — you only need to add what's new.
Guardrails
- This is about concision, not vagueness. A one-line prompt with no context
at all fails for the opposite reason — you still need to say what "good"
looks like if it isn't obvious.
- For genuinely ambiguous or high-stakes asks, more context is the right
call. The rule is to cut padding, not substance.
- If a short prompt keeps getting answers that miss the mark, that's a
signal to add one specific piece of context back, not to return to
over-explaining everything.
All 751 AI guides · JulieMango plans from £17/mo