01 of ten Working Notes on Prompting · What the Tool Is Actually Doing

What "You Are an Expert in X" Actually Does

A working note on a common AI prompting move, and what to do when you have actual domain knowledge.

You've probably been told that adding "you are an expert in [topic] with 20 years of experience" to a prompt will get you better answers. The framing does real work, but less than the advice suggests, and there's a more honest move available when you actually know the domain.

The common move

The familiar prompt You are an expert installation designer with 20 years of experience specializing in interactive media. Help me brainstorm an installation for a festival.

This pattern shows up in prompt engineering guides, in influencer content, in classroom advice. It feels like it's doing something powerful. The question worth asking: what is it actually doing?

What this framing does and doesn't do

What it does
  • Shifts vocabulary register. Pulls more domain-specific words into responses. Spatial dramaturgy, threshold experience, embodied interaction become more likely to appear.
  • Activates domain frameworks. Surfaces concepts from the actual discourse the model has read: presence theory, site-specificity, the audience-as-participant lineage.
  • Shifts the register of judgment. Produces more confident, opinionated responses about what's good or bad in the domain.
What it doesn't do
  • Doesn't add knowledge it lacks. The model has read about your topic either way. The framing changes retrieval and presentation, not knowledge.
  • Doesn't make outputs reliable. An expert-framed model fabricates specifics just as readily as an unframed one. Verification is still your job.
  • Doesn't substitute for your judgment. You're the actual expert in your practice. The framing gives you a more articulate sparring partner, not a replacement.

So the framing is a register-shift tool, not a knowledge-summoning tool. Used well, it changes how the model talks. It doesn't change what the model knows.

The more honest move when you have domain knowledge

When you actually know the field you're working in, claiming expertise for the AI is the wrong move. What you want is for the AI to engage you as a peer, not perform expertise at you.

A more honest framing For this work, write from the perspective of someone deeply familiar with [domain], particularly [sub-area you're working in]. Use domain-appropriate vocabulary. The person you're working with is [your actual relevant background, brief]. Treat my contributions as informed practitioner knowledge, not novice questions. Push back, challenge, bring real domain knowledge to the exchange.

This does more useful work than the generic expert claim because:

Try this

Next time you reach for "you are an expert in X," pause. Ask yourself: do I know this domain better than the AI does?

If yes, write peer-engagement framing. Tell the AI what register to use and tell it who you are. Ask it to push back.

If no, you don't actually need expert framing. You need to learn the basics first, then bring real questions. The AI's "expertise" can't substitute for understanding the field yourself.

The principle underneath

Prompt framings calibrate how an AI communicates with you. They don't summon capabilities it doesn't have. The best framings are honest about what you bring, what the AI brings, and what kind of exchange you want. The worst ones ask the AI to pretend, which always ends in slop.