PYMNTS tried to order food through ChatGPT. Get pizza from Little Caesars It took longer than driving to pick one up.
Starbucks And Little Caesars Fired Applications inside ChatGPT this month. Both brands are betting that consumers will order food in the same way they already plan meals and research products. The interface is familiar. Exit is not.
Preparation takes longer than ordering
Getting started takes work. Users go to the connector directory within ChatGPT, search for the brand and enable the integration manually. None of that happens in the original app. The Starbucks app opens with your saved preferences, loyalty balance, and reorder button. The ChatGPT path opens with an empty prompt.
PYMNTS asked for something low in fat and high in protein. Starbucks brought back a related drink, explained the option and allowed size selection within the chat. Nearby sites then appeared. The flow worked. Then the store was selected. A browser has opened. The Starbucks app took over. Log in, pay, check out. The conversation did not continue.
The Little Caesars never made it that far. The same question he asks for recipe suggestions and outside options. Nothing came from the actual list. The follow-up prompt produced one item. Then request the flow zip. No results were returned. The exit never happened.
Where Handoff breaks
Both brands are delivered to an external screen to complete payment. Starbucks routes the payment process through its app or website. Discovery occurs within ChatGPT. Payment is made outside. The company intentionally made this choice because its loyalty program is based on owning the transaction.
Advertisement: Scroll to continue
This division reintroduces the friction that the channel was designed to eliminate. The original Starbucks app completes the order in a second, keeping users’ saved payment credentials and loyalty status. The ChatGPT flow takes several minutes to reach the same point. Being delivered to the login screen doesn’t feel like a final step; It feels like starting over.
The back-end infrastructure is the problem. Payment credentials live in one system. Loyalty lives elsewhere. Order history lives elsewhere. The chat front end can’t connect to them. Users fill this gap themselves, by adding steps rather than removing them.
What is the measure of friction actually?
This is not a technological failure. The commercial infrastructure is not created for this purpose.
Bites builds its entire business model around ChatGPT as an order channel. Starbucks has intentionally kept the checkout process within its own ecosystem. None of the brands treat this as a gimmick. They are both making deliberate bets on which direction the interface will go.
These bets have a logic behind them. Conversational ordering removes the browsing and scrolling loop. A model that knows a user’s preferences, location, and time can outperform a static list. The discovery layer actually shows what that looks like.
The exit layer is not there yet. Credentials need to travel with the conversation. Loyalty status must persist across sessions. Payment must be completed within the chat. Until this infrastructure is in place, users will have to fill the gap themselves.
The tools work. Plumbing no. This gap will be closed. The question is whether consumers wait it out or go back to the three-click experience that already knows their order, address and card number. Habit is a short-term moat. As well as comfort. For now, the original app still offers both.
