Google is quietly reshaping what “Search history” means—expanding it beyond text queries into the messy, high-signal world of images, audio, and video. In a new update shared with users, the company says it will save the media you use when interacting with Search under a new control called “Search Services History.” The change matters because it touches exactly the kinds of inputs that modern AI-powered search relies on: a photo you point at something in the real world, a voice recording you generate while asking a question on the go, or a snippet of speech you feed into Translate.
For years, most people have thought of Search history as a list of searches—words typed into a box. But Google’s recent push into multimodal search has made that mental model outdated. Lens turns images into queries. Search Live turns spoken questions and live context into searchable information. Translate turns speech into text (and then into meaning). Now Google is explicitly grouping those media-based interactions into a single setting so users can see what’s being stored and decide whether it should be saved at all.
What’s changing, specifically
Google’s update describes a new “Search Services History” setting that governs whether certain types of media used in Search-related features are saved. The company indicates that this includes images, files, audio, and video associated with your Search activity. In practical terms, that means the media you provide to Search experiences—rather than only the results you click—can be retained.
The update calls out several categories of inputs:
1) Google Lens images
When you use Google Lens to search by image—uploading a photo or capturing one through your device—those images can be included in what Google saves under Search Services History.
2) Search Live recordings
Search Live is Google’s real-time search tool that uses audio and context to help answer questions as they happen. If you record or generate audio through Search Live, those recordings are part of the media Google says it may store.
3) Voice searches
Voice queries—spoken questions sent to Search—are also included. This is important because voice search often feels more “ephemeral” to users than typed text. You speak, the assistant responds, and you move on. But from a data perspective, voice is still an input that can be analyzed, transcribed, and used to improve systems.
4) Translate audio
Translate isn’t just about typing. When you speak into Translate, Google says the phrases you speak can be included in the saved media under Search Services History.
Google also notes that users can switch off the Search Services History setting. Additionally, there is a separate option to disable “Save Media,” which further controls whether these media items are stored. In other words, Google is giving users multiple layers of control, but the key point is that the company is making media retention explicit and configurable rather than leaving it implied.
Why Google is doing this now
This update isn’t happening in a vacuum. It arrives at a moment when search is becoming less like a website directory and more like an AI interface. The more Google leans on AI models to interpret images and audio, the more it needs training data and evaluation data. That doesn’t automatically mean every saved item is used for training, but it does mean the company wants the ability to review and learn from real-world interactions at scale.
There’s also a regulatory and trust dimension. Users increasingly expect transparency around data collection, especially for sensitive inputs like voice. By introducing a dedicated setting for Search Services History, Google is aligning its product behavior with a broader expectation: if you’re going to store media, you should make it easy to find and easy to turn off.
But there’s another subtle reason this matters: multimodal search blurs the boundary between “search” and “content.” A typed query is clearly a request. An image is content. Audio is content. Video is content. When those inputs are stored, they can reveal more than intent—they can reveal context, environment, and sometimes identity-adjacent details. A photo might include faces, documents, or location cues. A voice recording might capture background conversations. Even if Google processes these inputs carefully, the mere fact of storage changes how users perceive risk.
So the update can be read as both a technical adjustment and a communication strategy: Google is acknowledging that Search now involves media, and it’s offering a way to manage that media retention.
How “Search Services History” fits into Google’s broader privacy controls
Google’s privacy settings have historically been fragmented across different products and different types of data. Search history, Web & App Activity, Location History, YouTube watch history—each has its own logic and its own toggles. With Search Services History, Google is carving out a new category specifically for media used in Search-related features.
That design choice is telling. It suggests Google wants users to treat these media inputs as a distinct class of data—separate from general web browsing activity, and separate from other forms of account history. It also implies that the company expects users to care about media retention enough to warrant a dedicated setting.
The presence of both a “Search Services History” toggle and a “Save Media” option indicates Google is trying to give granular control. Users who want to keep some aspects of personalization might disable media saving while leaving other history features on. Conversely, users who want maximum privacy can turn off the entire Search Services History setting.
From a user experience standpoint, this is a meaningful improvement over older models where media retention might have been buried in policy language. From a trust standpoint, it’s also a signal that Google recognizes the sensitivity of storing images and audio.
What Google says it uses the saved history for
Google states that it will use Search Services History to provide and improve its Search services. That phrasing is common across tech companies, but it takes on extra weight here because the saved items are not just text logs—they are raw inputs that can be processed for transcription, interpretation, and pattern recognition.
In practice, saved media can help Google in several ways:
1) Improving recognition accuracy
Voice and audio inputs can be used to refine transcription quality, noise handling, accent robustness, and keyword detection.
2) Enhancing multimodal understanding
Lens images can help improve object recognition, OCR performance, and the ability to map visual cues to relevant search results.
3) Better evaluation of real-world performance
AI systems often fail in edge cases. Real user media provides a stream of edge cases that developers can test against.
4) Personalization and relevance tuning
Even when companies claim they use data primarily for improvement, saved history can also influence ranking and relevance. The line between “improvement” and “personalization” is often blurred in practice.
Google’s update frames the purpose as providing and improving Search features and services. It doesn’t present the saved media as purely archival. It positions it as part of the feedback loop that makes AI-driven search work better over time.
A unique angle: the shift from “what you searched” to “what you showed and said”
The biggest conceptual change in this update is that it reframes search history as something closer to “interaction history.” Instead of only tracking what you typed, Google is tracking what you displayed and what you spoke.
That shift has consequences for how users should think about privacy:
Typed text is relatively abstract. It’s usually short, intentional, and less likely to contain incidental personal details. Images and audio are different. They can include unintended information—documents you didn’t mean to share, people in the background, private conversations overheard by a microphone, or sensitive objects captured by a camera.
Even if Google applies safeguards, the risk profile changes when the system stores the raw media. That’s why the ability to disable saving is crucial. Without it, users would have little control over whether their most personal inputs become part of a dataset.
This update also highlights a broader industry trend: as AI becomes more capable, the “input layer” becomes more valuable. Companies don’t just want your clicks; they want your multimodal signals. And once those signals are stored, they can be used to train models, evaluate performance, and improve future interactions.
The trade-off users face: convenience versus control
There’s a reason Google is doing this. Multimodal search is powerful precisely because it can interpret messy real-world inputs. That power comes from learning from real interactions. If users opt out of saving media, Google may still process the input to answer the immediate question, but the long-term improvement loop could be reduced.
This doesn’t necessarily mean your experience will degrade immediately. Many systems can operate without storing raw media, or they can store it temporarily. But over time, the availability of saved media can affect how quickly a company improves recognition and understanding.
So the decision to disable Search Services History or “Save Media” is essentially a choice about how much you want to contribute to the improvement pipeline. For some users, that’s acceptable. For others, it’s not.
The update gives users a lever. It doesn’t remove the underlying reality that AI search benefits from data. But it does allow users to decide whether their images and audio should be retained.
What to do if you want more privacy
If you’re concerned about media retention, the update suggests two practical steps:
First, look for the “Search Services History” setting and consider switching it off if you don’t want Google to save images, audio, and video used in Search-related features.
Second, check whether there’s a “Save Media” option and disable it if you prefer not to store those media items even if other history features remain enabled.
Because Google’s settings can vary by region, account type, and product rollout timing, the exact path may differ. But the key is to treat this as a media retention control, not just a generic “history” toggle. The update is specifically about images, audio, and video used in Search experiences like Lens, Search Live, voice search, and Translate.
Why this matters beyond Google
Even if you don’t use Google Lens or Search Live, the implications extend to how AI systems are built and governed. This update is part