Telling an AI model that it’s an expert programmer makes it a worse programmer
Nobody is talking about this: telling an AI to act like an expert actually makes it worse at coding. Here's why your prompting strategy is backwards.
Why it matters
Research reveals that persona-based prompting degrades LLM performance on technical tasks—counterintuitive but critical for anyone building AI-augmented development tools or relying on prompt engineering as a deployment strategy.
The key facts
8 to knowPersona prompting (expert, specialist framing) reduces coding accuracy vs. neutral prompts
Finding challenges common assumption in AI product design and prompt optimization
Implications for coding assistants, IDE integrations, and enterprise AI deployment
Published March 24, 2026 — research-backed finding on LLM behavior
Persona-based prompting (e.g., 'You are an expert programmer') reduces model coding performance
Finding contradicts widespread industry practice in prompt engineering
Research published March 24, 2026 via The Register
Implications for enterprise AI training and workflow design
Go to the source
The Register AI/MLgo.theregister.com