Viral Complaints Over McDonald's Drive-Thru Order Errors

The post spread fast. A viral thread claiming that native English speakers get incorrect orders at McDonald's drive-thrus up to two-thirds of the time drew thousands of responses from people who said they had experienced exactly the same thing, adding fresh pressure to a reliability problem the chain has been struggling to manage for years.

The two-thirds figure, striking on its own, lands harder against McDonald's own benchmarks. When the company ran its AI-powered Automated Order Taker system at 24 Illinois test locations, the technology scored accuracy in the low 80% range, well short of the 95% threshold McDonald's said it needed before any national rollout. The company publicly claimed the program hit roughly 85% accuracy overall. Neither number satisfied customers who kept posting their receipts and bags of wrong food online.

By June 2024, McDonald's ended its AI drive-thru partnership with IBM entirely. The videos that accelerated that decision were hard to ignore: one showed a customer unable to remove a Diet Coke that the system had picked up from an adjacent lane's order; another captured the AI adding nine sweet teas after the customer tried to correct a single wrong drink. A separate clip showed ice cream orders compounding to three cones when a customer tried to remove bacon the system had added unprompted.

Those failures were framed at the time as a technology problem. The current viral thread complicates that narrative. The complaints now are about straightforward human order-taking, at standard drive-thru speakers, from customers with no accent barrier. That distinction matters for crew members who absorb the frustration at the window: the noise environment, the pressure to cycle cars through in under three minutes, and the volume of customization requests the modern McDonald's menu demands all pile onto the same interaction point.

McDonald's is now reintroducing AI to its drive-thrus in 2026, this time through a Google Cloud-powered platform, with the chain betting that improved speech recognition and noise cancellation will close the accuracy gap that torched its IBM experiment. Whether the new system can outperform the two-thirds failure rate described in the viral thread is the question its engineers need to answer before the cameras come out again.