How to Opt Out of Google Search's New AI Data Training Feature

The way we search the internet has fundamentally changed over the past year. Instead of merely typing text queries into a search bar, millions of users now point their smartphone cameras at unknown plants, hum half-remembered songs into their microphones, or use Android's popular "Circle to Search" feature to instantly identify objects on their screens. To support this multimodal future, Google is rolling out a significant update to its account settings this week. The company is separating traditional text history from a new, dedicated category called "Search Services History," which is designed to handle the complex files associated with modern artificial intelligence interactions.[3][4]

Under the newly implemented policy, any media you upload during a search interaction—including photos sent to Google Lens, voice recordings, translated documents, and screenshots—will be saved directly to your Google account. While this feature genuinely makes it easier to revisit past visual or audio queries, Google's updated support documentation reveals a critical secondary purpose. The stored media will be actively used to train the company's artificial intelligence models. This marks a notable shift from previous eras of search, where uploaded images were typically processed ephemerally and discarded. Now, as the internet pivots toward AI-driven discovery, these multimedia files are treated as valuable training assets.[2][5]

According to the new terms of service, Google can retain this uploaded media for up to four years to improve its speech recognition, visual identification, and generative AI systems. For users who rely heavily on features like live translation or document analysis, this means their personal uploads could become part of the training data for future iterations of Gemini and other Google AI tools. While the company maintains that this data helps build more accurate and responsive technologies for everyone, privacy advocates emphasize that users should be fully aware of how their personal files are being repurposed behind the scenes.[1][2][6]

Fortunately, Google has made it straightforward to opt out of this data collection without losing access to the underlying AI features. The update is explicitly designed to give users granular control over their digital footprints, replacing the old, monolithic "Web & App Activity" toggle with specific switches for different types of data. This modular approach means you can continue using advanced features like Circle to Search or voice translation without being forced to donate your media to Google's training servers. It is a welcome move toward transparency, provided users know where to look.[3][4]

To disable the media training feature and reclaim your privacy, the process takes only a few taps. Users need to navigate to the "My Google Activity" dashboard, which can be accessed via a web browser on any smartphone or computer. From there, locate the newly added "Search Services History" section, which governs data across Search, Maps, Translate, and Google Lens. Unlike previous iterations of Google's privacy controls, this new dashboard clearly delineates between text-based queries and the richer, more personal media files that power modern AI interactions.[5]

Inside the Search Services History menu, users will find a specific sub-setting that controls whether uploaded photos, audio, and videos are saved to the account. By simply unchecking this box, you instruct Google to stop storing your media files entirely. This single action effectively removes your future uploads from the AI training pipeline. It is important to note that disabling this feature does not break the functionality of Google Lens or voice search; the services will still process your media in real-time to deliver an answer, but they will immediately discard the file rather than archiving it for model training.[1][2]

For those who still want the convenience of revisiting their recent visual searches but do not want their data stored indefinitely, Google offers a practical middle ground. Users can leave the media saving feature enabled but configure a strict auto-delete schedule. This setting automatically purges the data from Google's servers every 3, 18, or 36 months, depending on the user's preference. This hybrid approach allows consumers to maintain a short-term history of their multimodal searches while ensuring that their personal photos and voice clips do not linger in corporate databases for the maximum four-year window.[2][5]

Google's update is part of a much broader industry trend reshaping the consumer technology landscape. Throughout late 2025 and early 2026, major artificial intelligence developers—including OpenAI, Anthropic, and Google—transitioned to "opt-out" privacy models. Under this paradigm, consumer conversations, uploaded documents, and visual queries are used for model training by default unless the user actively intervenes to change their account settings. This shift has placed the burden of privacy squarely on the shoulders of the consumer, making digital literacy more crucial than ever.[6][7]

Despite this industry-wide shift toward default data harvesting, a significant knowledge gap remains among the general public. Recent privacy surveys indicate that only about 22 percent of users are actually aware that these opt-out settings exist across their favorite AI platforms. Many consumers still treat AI assistants and smart search tools like private, closed-door conversations, entirely unaware that their prompts and uploads are routinely routed to remote servers for processing, safety monitoring, and potential integration into next-generation language models. As artificial intelligence becomes deeply embedded in daily workflows, bridging this awareness gap is essential for maintaining consumer trust.[6]

The push for user empowerment and explicit data boundaries extends far beyond Google's search ecosystem. Over the past year, platforms ranging from LinkedIn and X to newsletter hosts like Substack have all introduced specific toggles that allow users to shield their posts, comments, and professional data from AI scraping bots. As the web increasingly becomes a training ground for machine learning, knowing how to navigate these disparate privacy dashboards has become a fundamental digital skill for anyone who creates, shares, or searches online.[7]

As search engines continue to evolve from simple text retrievers into complex, multimodal AI assistants, the volume and intimacy of the data we share with them will only increase. Features like visual search and voice translation offer undeniable utility, breaking down language barriers and making information more accessible than ever before. By taking a few minutes to navigate these new settings, users can confidently embrace the next generation of search technology while maintaining strict, personalized boundaries around their digital lives.[1][4]