Food delivery may be entering a new phase as AI tools become a way to find restaurants and place orders. Bites, a small Bay Area startup, is testing a model that routes some ChatGPT orders straight to restaurants. Its pitch puts pressure on the fees and customer relationships that have helped define established delivery apps.
A different route from a chat to a restaurant
Bites describes itself as “AI native,” though it also has a traditional app. Diners can order from participating restaurants in ChatGPT, with the order going directly to the restaurant rather than through a third-party delivery platform.
The company says it works with point-of-sale systems such as Toast and Square. Those systems process restaurant orders and hold digitized menus; after a point-of-sale company partners with Bites, restaurants can opt in with a button. The arrangement aims to connect a customer’s request to a restaurant without requiring that restaurant to rely on a large delivery marketplace for the transaction.
Bites says it charges a flat $1 surcharge per order, typically paid by the diner, although some restaurants cover it. By comparison, DoorDash fees for restaurants range anywhere from 15 to 30 percent per delivery order. On DoorDash, restaurants that pay more in commissions become more visible to diners, according to the source article.
The fee debate reaches the restaurant’s bottom line
Bites argues that delivery commissions can push restaurants to raise menu prices, leaving diners to absorb some of the cost. Its lower-fee model is built on the idea that the ordering infrastructure and familiarity with delivery that platforms invested in are no longer as difficult to establish as they once were.
Matt Maloney, who cofounded Bites and Grubhub, argues that the old pricing structure reflects an earlier period of investment. Delivery companies spent billions building online ordering infrastructure and bringing restaurants, diners, and drivers onto their platforms. He says that restaurants increasing delivery-app prices to offset fees is a holdover from that era.
The article’s own comparison gives one example of how prices can differ. The Verge ordered a pizza and beverage through ChatGPT using the Bites plugin for $56.73, including tip, tax, and delivery fees. The same order on DoorDash totaled $70.29, with higher service fees and the same pizza priced $11 more. That comparison describes one order; it does not establish what every customer would pay across either service.
A small startup tests a bigger shift
Bites is far smaller than DoorDash. The startup has 10 people and around 300 restaurants signed up in the Bay Area, while DoorDash processed 970 million orders in its second quarter this year and generated $4.5 billion in revenue.
Still, Bites has drawn attention from DoorDash. In August, restaurants received a DoorDash email warning that they might be listed on Bites without consent. The message raised concerns about whether businesses had agreed to Bites’ “terms, pricing structures, or service standards,” whether their menus and operating hours were accurate, and whether such listings could be illegal depending on the state. It also explained how to ask Bites for removal.
Bites founder and CEO Bala Subramaniam said the company heard from 14 restaurant partners that received the letter. Some businesses that were not listed on Bites got emails too; the company said some restaurants had appeared in old demos and acknowledged that some had requested removal. A handful of businesses, after learning about Bites through the warning, wanted to join.
For restaurant owner Jay Jayaraman, the platform has become a significant source of orders. He says that before joining Bites, 80 percent of orders came from DoorDash. He estimates that now 65 percent come through Bites, 25 percent through DoorDash, and the remainder through Uber Eats and Grubhub. Jayaraman says Bites orders bring higher margins, and he welcomes the shift away from DoorDash because it adds to his bottom line.
AI agents could change how platforms make money
The stakes extend beyond delivery fees. If more people use AI agents to make purchases instead of opening apps and websites, platforms could lose opportunities to show related products, promote deals, offer free trials, sell ads, or feature boosted listings. Those tools are part of how online services shape the shopping experience and generate revenue.
DoorDash, for example, also earns money from advertising: its ads business hit $1 billion in 2024. If orders increasingly start in an AI conversation and flow directly to restaurants, a delivery app may have less influence over what customers see and fewer chances to promote an item during the process.
Bites offers a concrete test of that change, but its small size leaves open how far the model can grow. For now, it shows how AI ordering could challenge the role of delivery marketplaces: a simpler route to a restaurant may appeal to businesses seeking lower costs, while established apps face questions about how to maintain their place in the transaction.