Promotion or substitution? The effect of AI assistants on physician performance in online healthcare platforms

Purpose Artificial intelligence (AI) is fundamentally reshaping healthcare delivery. Physician-specific AI assistants, trained on individual physicians’ medical data, now provide personalized medical advice on online healthcare platforms. While AI assistants improve the availability of physician expertise, they may substitute for cognitive tasks traditionally undertaken by physicians. Despite this tension, few studies have examined how AI assistants affect physician performance. This study aims to examine the causal impact of introducing AI assistants on physician consultation volume and how physicians’ existing resource configurations shape this impact. Design/methodology/approach This study leveraged the introduction of AI assistants on a leading online healthcare platform to establish a quasi-natural experimental setting. We constructed a panel dataset covering 3,668 physicians and employed a difference-in-differences model to identify the causal effect of AI assistants on physician performance. Findings The results suggest that the introduction of AI assistants significantly improves physician performance, with physicians experiencing a 3.6% increase in consultation volume. Moderation analyses reveal that the positive effect is weaker for high-reputation physicians and those with broader knowledge. Further heterogeneity analysis shows that the effect is invariant to physician gender but more pronounced among physicians in non-first-tier cities. Originality/value This study provides causal evidence of the impact of AI assistants on individual performance in an online healthcare setting. It further reveals how physicians’ existing resource configurations shape AI’s impact from a resource-based view and offers suggestive evidence on its promotion and substitution pathways.