AI Guides › Prompting

The Anti-Pattern: Prompting Like You're Placating A Toddler

By Nigel Guy · 2 min read

There's a recognisable style of prompt that over-explains, over-flatters, and over-warns: "You are a world-class expert. Please, if you can, try to write a short paragraph. Take your time and think carefully. Thank you so much." None of this is a constraint. It reads like coaxing a reluctant toddler through a task, on the theory that enough softening language will somehow produce a better result. It doesn't — it just spends your words on tone management instead of information.

The tell

You're in this pattern if your prompt contains any of the following doing no real work:

1. Flattery as setup ("You are a brilliant, world-class…") — it doesn't change the output; the actual constraint (expertise level, register, depth) needs to be named directly if it matters at all.
2. Hedged requests ("if you could maybe try to…") — say what you want plainly. Hedging doesn't make the request more polite to a system reading text; it just makes the actual ask harder to locate.
3. Reassurance padding ("take your time," "there's no rush") — there's no clock to manage here; this line is doing nothing.
4. Over-thanking before you've seen the output — thank it after, if you like, but a "thank you so much" at the start isn't a constraint either.

The mechanism

Reread the prompt and, for each sentence, ask: is this an instruction, context, or a constraint — or is it just tone? Delete anything that's purely tone. Then check what's left. Often the coaxing language was quietly standing in for a real requirement you never actually stated — "take your time and think carefully" usually means you wanted a reasoned answer, not a fast one. Say that instead: "show your reasoning before the conclusion."

What to skip

Skip assuming brusqueness is the fix either — a prompt stripped down to a terse command with no context isn't better, it's just missing the other kind of information (audience, format, constraint) that actually matters. The goal isn't fewer words for their own sake; it's replacing filler with substance.

Guardrails

• Cutting coaxing language doesn't by itself make a vague prompt clear — it just removes the padding. You still need the actual constraints in there.
• Tone in a prompt isn't meaningless in every case — if a specific register matters to the task itself (a customer-facing message, for instance), that's a real constraint, not coaxing. The line is whether the language changes the request or just softens how you're asking.
• This is about prompt-writing efficiency, not a claim about what politeness does or doesn't do more broadly — that's a separate conversation.

All 751 AI guides · JulieMango plans from £17/mo