The Verge tested the new integration — the bot conversation took longer than waiting in an actual line, and the final drink still had to be clarified with a human anyway. Classic "AI product nobody asked for."
What the feature actually does:
You open ChatGPT, drop in a selfie of your outfit, and ask the model to recommend a drink that matches your vibe. The concept sounds like a Black Mirror episode written for Gen Z, but in practice it played out more like an absurdist comedy.
What went wrong:
- **Speed:** The editor spent more time chatting with the bot than he would have standing in a real queue with an actual barista. The AI spent ages "thinking" about the vibe of his jacket, suggesting options, asking follow-up questions — long enough to have brewed coffee at home.
- **Broken flow:** At some point the bot just lost track of the ingredients, and the drink order still had to be verified and corrected by a human.
- **The point:** This is a textbook example of [AI slop](/en/news/ai-for-good-reminder) — a product built because they could, not because it solves any actual problem. Instead of tapping two buttons in the regular Starbucks app, you're invited to roleplay with a language model.
Brands seem to be in a "let's bolt a neural net onto the fridge" phase with zero thought given to UX. Right now this looks like a clunky layer bolted on top of something that already worked fine. The irony is that the AI supposedly saving us time is, in this case, just eating it — for the sake of questionable fun.
I'm still waiting for the AI that just knows I need a double espresso without a conversation about my sneaker colorway.
*P.S. ChatGPT does have some genuinely great integrations. The Spotify one, for instance, creates any playlist directly in your streaming profile. I used it recently to build a setlist from a Rolling Stones concert I went to in Tel Aviv back in 2014.*