Get better answers from Claude by making it interview you first

so... you ask Claude for something, and the draft comes back fine.

Fine is the problem.

It's generic. It reads like it was written for anyone. And it's easy to blame the model for that.

But the model wasn't the problem. My prompt was two lines. The task had maybe ten decisions inside it... and Claude guessed every single one.

Why? Because I never gave it the answers. They were in my head.

What to do

Add one line at the end of your prompt:

Before you write anything, interview me first. Use AskUserQuestion.

Instead of a draft, Claude comes back with a few questions. Tappable options, not homework.

What it looked like for real

The cold ask: same project, no interview line. Claude starts writing immediately.

I asked Claude to draft a blog walkthrough of my AI news digest system. Cold, it has to guess how my system works. It guesses confidently. It guesses wrong.

The only change: one line added to the end of the prompt — "interview me first."

With the interview line added, it didn't write a word first. It asked:

Claude asks four questions before writing a word. Tappable options, about ten seconds of work.

Three taps. Where the news comes in (Gmail). What the task produces (a digest in chat). How long it's been running (weeks).

Ten seconds of tapping. Every guess removed.

The gap

Side by side: the cold draft guesses, the interviewed draft uses my actual answers.

The cold draft describes some generic AI-news setup that isn't mine. The interviewed draft describes my actual system, because I handed it the three facts it was missing.

Same model. Same task. One line of difference.

Try it

Whatever you were about to ask Claude next... add "interview me first, use AskUserQuestion" at the end.