Confident nonsense: Google’s AI Overview offers explanations for made-up phrases

Google’s AI Overview feature is displaying a peculiar pattern of generating fictional explanations for made-up idioms, revealing both the creative and problematic aspects of AI-generated search results. When users search for nonsensical phrases like “A duckdog never blinks twice,” Google’s algorithm confidently produces detailed but entirely fabricated meanings and origin stories. This trend highlights the ongoing challenges with AI hallucination in search engines, where systems present invented information with the same confidence as factual content.

How it works: Users can trigger these AI fabrications by simply searching for a made-up idiom without explicitly asking for an explanation or backstory.

• Adding “meaning” at the end of the fictional phrase seems to increase the likelihood of Google generating a detailed explanation.
• The AI doesn’t recognize these phrases as fictional and instead creates plausible-sounding definitions and origins that appear authoritative.

The Duckdog experiment: A ZDNET writer tested the phenomenon with a colleague’s invented phrase about her dog.

• When searching “A duckdog never blinks twice,” Google’s AI immediately produced a confident explanation claiming it meant a duck-like dog is so focused it never blinks twice.
• Subsequent searches for the exact same phrase yielded completely different explanations, with one claiming it refers to something “so unusual or unbelievable that it’s almost impossible to accept.”

Why this matters: This trend exposes a fundamental weakness in Google’s AI Overview feature that could undermine user trust in search results.

• While AI summaries can provide convenient quick answers, this experiment demonstrates they can also present fictional information as factual with complete confidence.
• The issue echoes Google’s previous AI mishap from a year ago, when its system suggested dangerous recipes like “glue pizza” and “gasoline spaghetti.”

Between the lines: This phenomenon represents a classic example of AI hallucination, where large language models confidently generate plausible-sounding but entirely fictional content when faced with inputs outside their training data.

• These systems prioritize providing answers over admitting uncertainty, creating a potentially misleading user experience.