DoorDash Data Reveals AI Agents Failing to Gain Traction in Food Delivery

The promise of AI agents automating the checkout process is hitting a wall of consumer indifference in the food delivery sector. During a third-quarter earnings call, DoorDash co-founder and CEO Tony Xu revealed that third-party AI agents have generated only "several hundred orders" in total since their inception, according to reporting from Ars Technica and TechCrunch.

To put that figure in perspective, Ars Technica notes that DoorDash handled just shy of 1 billion orders in the third quarter of 2026 alone. TechCrunch adds that the company's 2025 report showed an average of 3 million deliveries daily. While DoorDash recently introduced a Model Context Protocol (MCP) connector to help early beta partners and corporate lunch organizers streamline ordering, the vast majority of users continue to prefer the company's own app.

Xu suggested that while AI agents may eventually prove useful for "long consideration purchases" or proactive background work, the simplicity of ordering food means consumers prefer direct interaction. He cautioned against being "delusional" regarding the current state of the technology, stating that operators should not "fight gravity."

Interestingly, while third-party agents like Muse, Instinct, and Fo are struggling, DoorDash's internal AI tools are seeing more success. The company's "Ask DoorDash" feature, launched in June, allows users to build carts via photos or vague descriptions. DoorDash claims these baskets are built five times faster, feature nearly 50% higher basket values, and include about 60% more unique items than traditional orders, according to Ars Technica.