It is at the heart of our approach by helping people discover content from the web. AI mode allows people to easily explore and take action. A deep information search in the model allows people to express what they are looking for with all the nuances and constraints, allowing them to reach the right web content in a variety of formats.
Lab testing
We’ve got feedback internally and from trustworthy testers, and they’ve found that AI modes are extremely useful.
We are currently expanding our testing with limited opt-in experiences in our lab. This experimental approach helps you learn what’s most useful and quickly improve with feedback from people who want to try it out.
The AI mode is rooted in core quality and ranking systems, and also uses a new approach with the ability to infer the model to improve factuality. We aim to show AI-powered responses as much as possible, but if you are not confident in their usefulness and quality, the response will be a set of web search results. Like early stage AI products, we don’t always do it right. For example, we aim for AI responses during searches to present information objectively based on what is available on the web, but some responses may unintentionally acquire personas or reflect a particular opinion.
This next test phase addresses these types of challenges and rapidly modify the user experience based on feedback. We are already working on new features and updates, including adding visual responses with images and videos, richer formats, new ways to get useful web content.
Starting today, you’ll be trying Google One AI Premium Subscribers first on Labs. I look forward to your feedback and look forward to it more!
