Google Expands AI Search, Uses User Media for Training

Google is expanding its AI capabilities within Search, introducing a redesigned search box and new AI information agents, while simultaneously updating its privacy policy to use user-uploaded media for AI model training unless users opt out [1, 2]. These changes aim to integrate AI more deeply into Google's core search experience, but raise new questions about user data privacy [1].

AI Integration in Search

Google showcased a reimagined search box at Google I/O 2026, designed to streamline the transition between AI Overviews—AI-generated summaries at the top of search results—and AI Mode, a chatbot-like search experience [2]. Powered by the new Gemini 3.5 Flash model, the updated search box expands for longer queries and offers an AI-powered autocomplete feature [2]. Users will "reliably" see AI Overviews for natural-language questions, and can jump directly into AI Mode by attaching documents, photos, videos, or Chrome tabs to their queries [2]. This updated search box is rolling out globally across desktop and mobile devices starting recently [2].

Further enhancing its AI capabilities, Google plans to introduce new AI information agents for AI Pro and AI Ultra subscribers this summer [2]. These agents can monitor topics offline, such as tracking a band's tour dates, and notify users of updates [2]. They can also connect to other Google accounts via Personal Intelligence to pull information from apps like Gmail for personalized responses [2]. Google is also expanding its AI-powered booking feature this summer, enabling it to schedule local experiences and services, and even call businesses like pet groomers or home repair services on a user's behalf [2]. In the coming months, a new "generative UI" feature will allow users to build custom widgets and apps directly within the search engine, with AI Overviews capable of coding interactive visualizations [2]. AI Pro and AI Ultra subscribers will also gain access to a feature allowing them to create "super apps" in Search, such as a home fitness tracker connected to real-time information like weather and calendars [2].

Data Use for AI Training

Alongside these feature enhancements, Google has updated its search privacy settings, automatically opting all users into expanded AI training [1]. This new policy means that any media generated by user searches or uploaded to Google's services—including images, files, audio, and video from interactions with Search services—may be used to develop and improve Google’s AI models and technologies [1]. This includes media from screenshots taken with Circle to Search, uploaded photos, or voice searches via Search Live or Google Translate [1].

Users can opt out of this feature to prevent their media from being used for AI training [1]. This can be done through the Search Service History and Search Service Personalization pages within the Google app settings, or by visiting myactivity.google.com and unchecking "Save media" under Search Services History [1]. Users can also configure how long saved data is retained, with options for 3, 18, or 36 months [1]. Disabling personalized ads at myadcenter.google.com is another privacy option [1].

What to watch

• Global rollout of AI Search features: Monitor the adoption and user feedback on the redesigned search box, AI Overviews, and AI Mode as they roll out globally [2].
• Subscriber uptake of AI agents: Observe the impact and usage of new AI information agents and "super app" creation for AI Pro and AI Ultra subscribers this summer [2].
• User response to privacy policy changes: Track user opt-out rates and public discourse regarding Google's expanded use of user media for AI training [1].

Google's dual approach of significantly enhancing AI capabilities in Search while expanding its data collection for AI training presents a trade-off between advanced functionality and user privacy, with the long-term implications for both user trust and AI model development remaining to be seen.