Google's New AI Experiment Draws Backlash: Misleading Headlines, Fewer Clicks
Photo by Alex Dudar on Unsplash
Source
engadget.com
Published
Dec 3, 2025
Category
AI & Machine Learning
Executive summary
Google used artificial intelligence to generate the titles of some articles on Discover. This resulted in the content and the title being completely different. The title is entirely incorrect or misleading. Google explained itself as follows: “An interface experiment for a small portion of Discover users.” The explanation was not very convincing. Google is testing an integrated AI experience on a single screen by combining AI Mode with AI Overview.
Analyst insight
The direct use of artificial intelligence in critical content such as articles can lead to erroneous results; therefore, it is essential for the output generated by AI to undergo a final human review for both accuracy and trustworthiness. However, Google's increasing integration of AI Mode into its Search platform indicates that the company is moving away from its traditional search model. This makes the results more visible despite AI's error-prone nature, creating risks for users and leading to fewer clicks for publishers, making the process more problematic. As a result, the risks for publishers increase, bringing with them risks such as traffic loss and revenue loss. Google's stance undermines the integrity of the internet ecosystem. Google's vision of “AI being everywhere” is not yet ready for implementation.
Startups & BusinessFormer Chancellor of the Exchequer George Osborne is now part of the OpenAI team. When Osborne asked himself which company would be the best in the near future, he found the answer in OpenAI. He will also be on the board of directors for the “OpenAI for Countries” program, which supports countries in their adoption of artificial intelligence. With Osborne joining OpenAI, the aim is to work with governments to improve public services.
Mobile & DevicesCopilot is now available on LG smart TVs. Microsoft's AI chatbot comes pre-installed on computers and other technologies and cannot be uninstalled. However, this is related to the privacy and permission settings you share with LG.
AI & Machine LearningElon Musk's X platform's artificial intelligence, Grok, repeatedly provided incorrect information about the Bondi Beach attack. Grok stated in posts that the hero who neutralized the attacker was not real, but it has begun to correct its mistakes. The reason for the error stems from the fact that the news Grok referenced was generated by artificial intelligence.