Coaching or Crutching? The Impact of AI Collaboration Modalities on Job Interview Performance

The rapid adoption of AI tools has fundamentally transformed how job candidates prepare for interviews. While prior research highlights AI’s productivity benefits, little is known about how different AI collaboration modalities influence interview outcomes. This study examines two distinct roles of AI: Crutch AI (reactive information provider) and Coach AI (proactive guidance and feedback provider). In a controlled experiment with 200 job-seeking participants, we compared subjective and objective interview performance across an analytical coding task. Results showed that participants collaborating with Coach AI reported significantly higher confidence in performance and produced code with greater readability. However, external evaluators rated the functionality of their code as not significantly different from those who used Crutch AI. These findings suggest that Coach AI may induce a productivity illusion, artificially inflating users' confidence in their performance without yielding actual improvements in objective outcomes. This study contributes to human–AI collaboration research and offers practical implications for the design of professional AI.