DoorDash Ask chatbot launches with text and photo ordering

DoorDash rolled out Ask DoorDash, a new AI chatbot that lets users describe meals in natural language—"something spicy under $15," "low-carb lunch near me"—or upload a pantry photo to get restaurant suggestions and a ready-to-checkout cart.

The feature went live this week across iOS and Android. Instead of scrolling through restaurants and categories, users type or speak requests, and the system surfaces matching items, builds a cart, and routes the order. The chatbot retains context across a session; asking "make it vegetarian" after an initial query adjusts the cart rather than starting over. The feature appears in the search bar at the top of the home screen, and users can toggle between traditional browsing and the conversational interface.

DoorDash has not disclosed which language model powers the chatbot or whether the image-recognition feature uses a multimodal foundation model. The photo-upload capability suggests a vision-language architecture similar to GPT-4V, Claude 3, or Gemini Pro Vision, though the company may be running a fine-tuned open-weight model internally. The lack of technical disclosure is typical for consumer-facing AI features at scale—vendors rarely surface the underlying stack when the product is the interface itself.

The rollout follows similar conversational-ordering experiments at Instacart and Uber Eats, both of which added AI search bars in the past year. Those implementations have leaned heavily on retrieval-augmented generation to ground responses in real-time inventory and menu data, a pattern that Ask DoorDash likely replicates. The challenge for all three platforms is the same: natural-language queries are ambiguous, and users expect the system to infer preferences—dietary restrictions, price sensitivity, delivery speed—without explicit filters.