Evidence45%Reported
FactConfirmedProduct·April 9, 2026
Oumi Study: Google AI Overviews Generate Millions of Inaccurate Answers Per Hour
A study by startup Oumi found that Google's AI Overviews powered by Gemini 2 and Gemini 3 were accurate only 85% and 91% of the time respectively, meaning hundreds of thousands of false answers per minute given Google's search volume, drawing criticism from the News/Media Alliance.
Evidence Strength
Evidence45%Reported
Based on original reporting
Single publisher source
Insights
First tracked
April 9, 2026
Last updated
April 9, 2026
Sources
1 source
Related Developments
Meta Launches Muse Spark AI Model to Compete with Google Gemini and OpenAIGoogle Maps Rolls Out Gemini-Powered AI Photo Captions for ContributorsGemini Launches "Notebooks" Feature for Organizing ProjectsGoogle Integrates NotebookLM Fully into Gemini AppGemini AI Gains Interactive 3D Model and Simulation Generation Capability
Sources (1)
Source Timeline
Evidence Strength
Evidence45%Reported
Based on original reporting
Single publisher source
Insights
First tracked
April 9, 2026
Last updated
April 9, 2026
Sources
1 source
Related Developments
Meta Launches Muse Spark AI Model to Compete with Google Gemini and OpenAIGoogle Maps Rolls Out Gemini-Powered AI Photo Captions for ContributorsGemini Launches "Notebooks" Feature for Organizing ProjectsGoogle Integrates NotebookLM Fully into Gemini AppGemini AI Gains Interactive 3D Model and Simulation Generation Capability
