DoorDash Warns Restaurants Over ChatGPT Ordering Startup Bites
DoorDash emailed Bay Area restaurants in August warning that they may be listed on Bites without their consent.
Bites is a 10-person, pre-seed startup at the very early funding stage, with around 300 Bay Area restaurants signed up. It lets diners place orders from those restaurants directly in ChatGPT, with orders going directly to the restaurant. The Verge
DoorDash said it had heard from several partners that they were added to Bites without their knowledge. The email said those restaurants may not have agreed to Bites' terms, pricing structures, or service standards. They may also lack control over menu or hours accuracy on Bites. It added that depending on the state, such listing practices could be illegal. It included instructions on how to contact Bites to request removal.
The dispute centers on take rate and order routing. Take rate is the commission cut taken on each order. DoorDash charges restaurants 15% to 30% commission per delivery order. Bites charges a flat $1 surcharge per order. Bites describes itself on its website at withbites.com as a fair, AI-native food ordering network with no menu markups, no hidden fees and no commissions, and states that diners get the same menu prices as ordering direct. AI-native means built around AI from the start. On LinkedIn, it describes itself as an AI-native marketplace combining the convenience of third-party delivery with the economics of first-party, or restaurant-run, ordering.
Bites founder and CEO is Bala Subramaniam. The company said it heard from 14 restaurant partners who received DoorDash's warning letter. One operator, Jay Jayaraman, has 13 restaurants listed on Bites.
Scale explains the sensitivity. DoorDash processed 970 million orders in its second quarter, generating $4.5 billion in revenue. The Verge Those results were up 27% and 36% year over year, respectively, in Q2 2026, with adjusted EBITDA of $914 million. DoorDash Adjusted EBITDA is a common measure of operating profit. The company trades on NASDAQ under the ticker DASH, was founded in 2013, has expanded to over 30 countries, and operates through its Marketplace and its Commerce Platform.
DoorDash is also building its own AI surface. In December 2025 it launched Zesty, an AI-powered app for finding new restaurants that provides AI-powered restaurant recommendations and lets users browse posts made by others. Bloomberg In March 2026, the company began paying delivery couriers to submit video clips and complete other digital tasks to help improve artificial intelligence.
The broader context here is the shift from marketplace search to agentic ordering, where an AI assistant completes the purchase inside the chat. If a diner can complete a transaction inside a conversational model, the discovery layer, the checkout logic, and the system of record for menu, hours, and pricing can separate. For operators, that separation creates two distinct risks: loss of control over data accuracy, and uncertainty over who owns fulfillment, refunds, and support when the storefront is an LLM, a large language model like ChatGPT. For incumbents, the risk is losing high-intent traffic without a corresponding cut of the ticket.
In my view, the economics will decide this faster than the legal arguments over unsolicited listings. A $1 surcharge against a 15% to 30% commission is a stark comparison for independent restaurants operating on thin margins, particularly if first-party fulfillment holds up. Worth flagging, lower distribution costs have historically expanded order volume rather than simply shifting it, from the PC web to mobile to marketplaces. If agentic checkout proves reliable on accuracy, latency, and exception handling, it could pull more direct orders into existence instead of only rerouting existing ones.
