Have you ever taken a photo of a swimsuit with Google Lens to find out where it was from? Or entered a voice query? We have some news you may not have heard about. From now on, Google stores all this data and automatically uses it to train its artificial intelligence, Gemini, unless you manually turn this feature off. This was reported by TechCrunch and Euronews. The good news, however, is that there is a solution.
What Google learns from your data
Google announced that Gemini would automatically begin learning from your data rather quietly. Users received an email about an update to their privacy settings, and most people did not pay much attention to it. However, the message announced the introduction of a new setting called Search Services History, which is enabled automatically for most accounts.
And what exactly counts as training data? According to data analyst Annie Cushing, who writes about the topic on Medium, it includes more data than we might expect. This includes not only your search queries and information from pages visited through Google, but also conversations with AI, your approximate location, and media, meaning photos uploaded through Google Lens. But it does not end there. Voice queries and data from Translate, Maps, Shopping, and other services are also under scrutiny.
However, Google Photos and YouTube are explicitly excluded from the new policy. According to Google, the new update will not affect data from these services.
Why is Google doing this, and how long will it keep the data?
Google uses your search and activity history not only to improve its services but also to train AI. According to available information, it may retain the data for up to 4 years. Experts therefore recommend considering what you enter into these services before continuing to use them, including biometric data such as voice recordings or facial images.
Google is not alone
According to Euronews, however, this is part of a broader trend. Google is certainly not the first company to use not only data from the web but also user-generated content. One example is Meta, which began training AI on public posts by European Facebook and Instagram users in 2025 and is now even facing a lawsuit for reviewing footage from its users’ smart glasses.
Similar criticism was also directed at Anthropic, which developed the Claude chatbot and prided itself on its reputation as a company that did not use user data to train AI. However, in August 2025, it changed its terms so that using conversations to train AI became the default setting. It obtained consent through a pop-up window whose design critics described as a so-called dark pattern, meaning a manipulative interface.
How to prevent Google from accessing your data
If you do not want Google to store what you search for, simply visit myactivity.google.com/activitycontrols, sign in, and turn off all history under Search Services History. However, because the company is rolling out the change gradually, the new setting may not appear for you yet. In that case, simply turn off Web & App Activity. Google has confirmed that it will carry this choice over to the new setting once it becomes available to users.
Sources: Euronews, TechCrunch, Medium, Euronews, The Decoder, Google Help
