The Bug Reports Are Rolling In

On June 25, 2026, a Hacker News user posted what might be the most surreal SEO nightmare Google has faced in recent memory: searching for "keynesian economics" on Google's AI Overview feature returned results where the first paragraph appeared in English—but every subsequent section was rendered entirely in Korean. The kicker? The user claims they have never installed or used Korean language settings on any of their devices.

Reproducible Across Multiple Environments

What's particularly concerning about this incident is that it wasn't an isolated browser quirk. The original poster confirmed the behavior was reproducible across multiple browsers and devices, which suggests something deeper than a local cache issue or extension conflict. This points to either a server-side rendering bug within Google's AI Overview pipeline or potentially a localization routing error that's being triggered by certain query patterns.

The Question Window Workaround

One interesting detail: when the user submitted an actual question to the AI through the interface, it opened a new window—and that new window rendered everything in proper English. This suggests the core language model is functioning correctly and has not mysteriously developed multilingual schizophrenia. Instead, something in the initial overview rendering layer appears to be malfunctioning for specific query contexts.

Could This Be Related to Training Data Contamination?

For those who follow AI systems closely, this raises some intriguing possibilities. AI Overview likely pulls from multiple sources to construct its summaries—including academic papers, cached web content, and potentially multilingual training data. If a particular topic like Keynesian economics has strong representation in Korean-language academic databases or cached documents that Google indexes, the system might be incorrectly weighting that language variant during synthesis.

Localization Bugs at Scale Are No Laughing Matter

While this instance appears to be more amusing than harmful, localization bugs at Google's scale can have real consequences. Users relying on AI Overview for accurate information—students researching economic theory, professionals looking up technical concepts—could walk away with fundamentally incorrect assumptions about what language the source material was originally written in.

Key Takeaways

  • Google AI Overview appears to mix languages within a single query result under specific conditions
  • The bug is reproducible across different browsers and devices, suggesting server-side causes
  • Submitting direct questions to the AI bypasses the issue entirely
  • This could indicate problems with source weighting during multilingual content synthesis

The Bottom Line

Google's been their AI Overview as the future of search, but bugs like this show how fragile these systems are when they start synthesizing content from multilingual sources. Until they nail down why certain queries trigger language bleed-through, I'd recommend double-checking any technical or academic topics through traditional search—especially if you don't read Korean.