Decoder. plain-English AI glossary

Persona prompting

● Core

Explicitly telling the model to adopt a character or perspective — doctor, poet, skeptic — to shape reasoning and tone.

Think of it like

Method acting: the model immerses itself in a character.

Example

"You are a seasoned detective. Analyze this crime scene." vs. just "Analyze this crime scene."

How it actually works

Personas work by leveraging associations learned during training. A "data scientist" persona elicits analytical language. A "therapist" persona elicits empathy. Personas can be combined with other techniques. Over-specific personas ("a 400-year-old vampire who codes in Rust") may cause hallucination.

For product teams

Inexpensive way to steer behavior and tone.

For engineers

Test personas for consistency. Avoid hallucination in overly specific personas.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome