Skip to content
Submit incident
AII-1001-87611
ReportedReported by fewer than three independent publishers, with no official finding.Supporting sources are pending verification.

Google AI presents April Fools' prank about Welsh roundabouts as factual news

A Welsh journalist's five-year-old April Fools' Day prank claiming his town set a record for roundabouts was presented as fact by Google's AI Overview tool in 2025. The AI quoted the fabricated figures and fake resident quotes, demonstrating the system's inability to distinguish between intentional satire and real news, particularly when the source lacked the original disclaimer.

1 April 2025 Β· Google Β· Reliability and hallucination
1 April 2025
AII-1001-87611
Quick facts
Category
Reliability and hallucination
Country
United Kingdom
Date
1 April 2025
Drafted by
Raidu Incident Detection

What happened

Journalist Ben Black published an April Fools' Day story on 1 April 2020 on his community news site Cwmbran Life claiming that Cwmbran had set a Guinness World Record for having the most roundabouts per square kilometre. Black added a disclaimer that afternoon clarifying it was an April Fools' prank.

On 1 April 2025, five years later, Black searched for his old story and found it being presented by Google's AI Overview tool as factual information, with the AI quoting the made-up calculations and fake resident quotes from the article. The original story had been republished by a larger national news website without Black's permission, and despite his efforts to have it removed, it remained online.

Reported impact

Black said he was 'shocked' and 'worried' to find the false information being used by Google's AI tool and presented as real information.

Index analysis

The AI Overview pulled information from a republication of the story that did not include Black's April Fools' Day disclaimer, even though he had added one to the original article on the day of publication. AI systems do not have the ability to understand the difference between intentionally fictional content and factual reporting.

Classification

Category
Reliability and hallucination
Industry
Media, social platforms and marketing
Country
United Kingdom
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Public interest
Harmed
General public
Failure modes
Agent misaction, Hallucination, Misidentification
AI systems
Google AI Overview

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident