Does Google Photos use your photos to train AI? What Google's pages say
The Morse teamPublished 11 min read
As of October 2026, Google says it doesn't use your photos to train generative AI models outside of Google Photos, and that promise leaves three things open. It says nothing against models inside Photos, and Google's own privacy hub says it "may process your photos and videos to improve edits or make inferences." People at Google may read the questions you type into Ask Photos unless you switch that off. And once you connect Photos to Gemini, Gemini's rules apply, which let Google train on "summaries, inferences, and generated media" drawn from your library.
We read Google's own pages on October 27, 2026: the Gemini features in Photos privacy hub, the Ask Photos, Face Groups, editing, and AI creation help pages, Google's Photos privacy page, the Gemini help pages on connecting Google Photos, the Google Privacy Policy (effective October 16, 2026), and the Google Terms of Service (effective July 30, 2026). Every fact below links to the Google page it came from. Google's Photos privacy hub says reviewers who read your Ask Photos questions never get your photos with them, and it lists every item that goes into a feedback report.
Google Photos AI settings at a glance
Read on October 27, 2026. Paths are for the Android app.
| Setting | Default | How long Google keeps it | Human review | Source |
|---|---|---|---|---|
| Training generative AI outside Photos | No switch; Google says it doesn't | Not applicable | Not applicable | |
| Use Gemini in Photos and Ask Photos | Off until you opt in | Remember List facts stay until you delete them | Answers only if you send feedback, or for abuse or harm | |
| Allow us to access your queries | Google calls turning it off opting out | Not stated | Your Ask Photos questions, cut from your account first | |
| Your Ask Photos Activity (feedback) | Keeps the feedback you send | Until you delete it; "deletion isn't immediate" | Feedback reports, with a transcript and lists of photos used | |
| Face Groups | Not stated on Google's page | Until you turn it off, or two years without using Photos | None described | |
| Estimate missing locations | Not stated | Past estimates stay after you turn it off | None described | |
| Google Photos in Gemini (Connected Apps) | Off until you connect it | Kept in Gemini Apps Activity, even after you disconnect | Media only for abuse or harm, or feedback you send |
Does Google train AI on your Google Photos?
Not generative models outside Photos, Google says. The Gemini features in Photos privacy hub states: "We don't train any generative AI models outside of Google Photos with your personal data in Google Photos." Google's Photos privacy page adds that Photos "doesn't sell your photos to anyone" and "doesn't use your photos and videos for ads."
That promise covers only generative models outside Photos. The same hub says Gemini features use your library, your face group names, and your Google Account information, and that Google "may process your photos and videos to improve edits or make inferences." Its examples of inferences are "guessing the age and locations of your top face groups and inferring insights related to your life." The hub doesn't say whether Photos' own models are trained on your library, and we found no Google page that does.
Two broader Google documents also apply. The Google Terms of Service let automated systems analyze your content "to recognize patterns in data, such as determining when to suggest a new album in Google Photos," and say this happens "as the content is sent, received, and when it is stored." The Google Privacy Policy gives "publicly available information" as its example of what Google uses "to help train Google's AI models," and lists photos you save among the content it collects. For how Google's rules compare with other companies', see which AI companies train on your data.
Can people at Google read your Ask Photos questions?
Yes, unless you turn it off. Google's Ask Photos help page says: "To help us improve Photos, Ask Photos queries may be reviewed by humans, but we first take steps to protect your privacy." The hub lists those steps as "disconnecting queries from Google Accounts and using specialized processes to remove sensitive content," and says a reviewed query doesn't include "the Ask Photos response, your photos or videos, or any information from your account."
A question can identify you by itself. "When did Maya and I move into the Elm Street apartment?" carries names and an address with or without your account attached. To stop review, open Google Photos, tap your profile picture, then Photos settings, Preferences, Gemini features in Photos, and Help us improve photos, and turn off Allow us to access your queries. Google titles the step "Opt-out of query donation." BGR's May 27, 2026 guide to avoiding Photos' AI features covers the Gemini toggles and leaves this switch out.
Feedback opens more. When you tap thumbs up or down, the privacy hub says the report carries:
- The conversation: your questions and Ask Photos' answers.
- An optional screenshot, which may show your photos.
- The photos it studied, listed with "the timestamp, location, faces in the photo, and sometimes AI-generated text descriptions," but not the pixels.
- The albums and face groups it studied, including name labels, estimated details about each person, and facts from your Remember List such as "their birth month or year, or their gender."
Feedback is stored in Photos Activity at myactivity.google.com/product/photos, where you can delete it. Google says "deletion isn't immediate," and that if you marked a query as abusive and asked to be contacted, Google Photos "might still contact you even if you delete the feedback."
How do you turn off AI in Google Photos?
Each feature has its own switch, and most sit in one place. In the Google Photos app, tap your profile picture, then Photos settings, Preferences, and Gemini features in Photos (Google):
- Turn off Search with Ask Photos to go back to classic search.
- Turn off Use Gemini in Photos to stop Ask Photos and the other Gemini features together.
- Open Remember List and delete the facts it saved about you and the people in your photos.
- Turn off Your Ask Photos Activity to stop saving feedback, then delete what's there.
- Turn off Help me title, which uses generative AI to suggest titles for your memories (Google).
Two more switches feed every Gemini feature. Face Groups is under Photos settings, then Privacy (Google). Estimate missing locations is under Settings, then Location on photos.google.com, and turning it off leaves past estimates in place until you edit or delete them (Google).
What does Face Groups keep about your face?
A numeric model of every face it groups. Google's Face Groups page says algorithms "create face models that numerically represent the images of faces," and that "these models may be considered biometric data in some jurisdictions." Photos may also group a person by "photos being taken close together in time" and "the same clothing across photos when a face isn't visible."
Your labels stay private. Google says face groups and labels "are only visible to you" by default and aren't shared when you share photos. The models last until you delete them or your account "is inactive for more than two years." Turning Face Groups off deletes your face groups, face models, and labels, and deleting one photo deletes the face models made from it (Google).
Face Groups "isn't available in all geographic regions, all domains, or all account types," and Google's page doesn't list where. Every Gemini feature in Photos requires it, with your own face picked as "me" (Google).
Does Gemini see your Google Photos?
Only if you connect them, and then it can see a lot. Google's page on Photos in Gemini says you need to be 18 or over, on a personal account, with Face Groups on, your face picked, and Estimate missing locations on. It adds: "Gemini can't connect to gallery apps" when Keep Activity is off, so connecting Photos means Gemini saves your chats. One of Google's example prompts is "@Google Photos what's my driver's license number?"
For the photos themselves, Gemini's Connected Apps page says Google doesn't train "on imagery or audio from your Google Photos library," and that reviewers see the media only "for cases of abuse and harm or when you submit feedback." What Gemini does with everything it draws from your connected apps, including training on what it infers, is in is Gemini private.
Personal Intelligence, which links Photos to Gemini with your other Google apps, isn't offered in the European Economic Area, Nigeria, Switzerland, or the United Kingdom (Google), and connecting Photos "isn't available in some United States regions" (Google). Gemini can now also run tasks on your library, such as curating albums and editing copies of photos, and Google says it asks before sharing an album or sending an email (Google).
Sharing a single photo out of Photos is different again. The privacy hub says a photo you upload to Gemini follows the Gemini Apps Privacy Notice, and one you search with Lens may become part of your Search history, "which may be used for ads." Our page on Google AI Mode's privacy covers that history.
Every AI feature in Google Photos, and who can review it
Read on October 27, 2026, from Google's help pages and its November 11, 2025 launch post. "Needs" lists the settings each feature requires.
| Feature | Needs | What Google keeps | Who can review it | Source |
|---|---|---|---|---|
| Face Groups | Nothing extra | Face models, groups, and labels until you delete them or two years idle | No review described | |
| Ask Photos questions | 18+, Face Groups, "me" picked, location estimates | Questions it may pass to reviewers; for voice, only that you asked by voice | People, unless you opt out | |
| Ask Photos answers | Same | Remember List facts you or it saves | People only with feedback, or for abuse or harm | |
| Ask Photos feedback | Same | Transcript, photo lists, face group details, in Photos Activity | People, by design | |
| Help me edit and Ask edits | 18+, personal account, Face Groups, Gemini features on | The edited photo, labeled "Edited with AI tools" | Feedback you send | |
| Nano Banana restyle | Same as Help me edit | The restyled photo | Feedback you send | |
| Photo remix and Video remix | 18+, in one of 12 listed countries | What you save; a daily generation limit | Feedback you send | |
| Personalized templates | Gemini features in Photos | Uses "insights from your life" | Feedback you send | |
| Help me title | 18+, mostly US | Suggested titles on your memories | Feedback you send | |
| Photos in Gemini | Keep Activity on, Face Groups, location estimates | Gemini chats, under Gemini's rules | Reviewers under Gemini's rules; media only for abuse or harm, or feedback |
The 12 countries for Photo remix and Video remix, as of October 27, 2026, are Argentina, Bangladesh, Brazil, Colombia, Egypt, India, Indonesia, Japan, Mexico, Pakistan, the Philippines, and Turkey. Help me edit fixes faces using "images from your private face groups," Google's November 2025 post says. When you save an AI edit, Google Photos adds Content Credentials to the file, and its "How this was made" panel shows "Edited with AI tools" (Google). For the editing and creation features, Google's help pages describe no review beyond the feedback you send and make no training statement of their own, so the hub's promise about "your personal data in Google Photos" is the one that applies.
How Morse keeps your questions about your life private
Every Gemini feature in Photos asks for more of your life before it works: Face Groups on, your own face picked, your locations estimated, and, for Gemini, your chats saved. The questions you ask about it then sit in the same Google Account as your photos, where reviewers may read them and Gemini may train on what it infers. Morse is built so your history doesn't sit on our servers. Your chat history lives in your browser, encrypted, and Morse doesn't train AI on your content, sell your data, or show ads (Morse privacy policy).
Ask Photos invites questions about your own life, like Google's example "What did I eat on my trip to Barcelona?" Many of them you can ask in words instead, with only the details you choose to type: planning a trip like the last one, a toast for a parent's birthday, or how to talk to a friend after a falling out. There's no query review switch to find, because Morse's servers don't keep your messages or log what's in them.
Each chat has its own privacy mode:
- E2EE, for the questions you'd put to Ask Photos about the people you love. Before your words leave the device, your browser checks that the secure enclave is genuine and seals the message for that enclave alone, so what passes through our servers is unreadable to us. It answers with open-weight models and has no web search or image generation, and the guarantee depends on the enclave hardware and attestation working as designed. It's part of Pro, $19 a month, and our guide to end-to-end encrypted AI chat explains the checks.
- Private, for planning the next trip or the next birthday. Every new chat starts in it. DeepSeek, Kimi, GLM, and Grok answer from hosts that agreed to keep nothing. When a provider confirms zero retention reply by reply, Morse verifies that and turns away any reply without it.
- Anonymous, for Gemini with no Google Account attached. The message reaches Google under Morse's account, with your account, email, and IP address stripped first. Google reads the text without knowing it's you and may hold it for a limited time to check for abuse, under its own policy, so write "my sister" rather than her name.
In every mode, your history locks when you sign out and opens with two keys, one of which never leaves your device. Morse's Studio, its image generator, keeps the images you make in your browser too, encrypted. Clearing your browser deletes your history for good, and Morse can't recover it, so copy anything you'll want later. Because we don't store chats, we usually have very little to give authorities.
What to do, by concern
- Training on your library: don't connect Google Photos to Gemini or Search if you don't want summaries and inferences from it used for training. Morse doesn't train AI on anything you write.
- Reviewers reading your questions: turn off Allow us to access your queries, and don't send feedback on Ask Photos answers about other people. On Morse, an E2EE chat reaches only the verified enclave.
- Your face: turn off Face Groups to delete your face models, knowing every Gemini feature in Photos requires it.
- What Gemini already saw: delete it in Gemini Apps Activity, using our steps to delete Gemini history.
Try a chat from the Morse home page without an account, or start free with no card needed.
Common questions
Can Google employees see my Google Photos?
Not as a matter of course. Google's Gemini Connected Apps page says that even with Photos connected, "except for cases of abuse and harm or when you submit feedback, humans don't view media from your Google Photos library." Google's Terms of Service also say automated systems scan stored content for spam, malware, and illegal content.
Does deleting a photo delete its face model?
Yes. Google's Face Groups retention policy says deleting a photo or video deletes "any face models derived specifically from that photo or video." Deleted items sit in your trash for 30 days before they're permanently deleted, so empty the trash if you want them gone sooner.
Does Google Photos use my photos for ads?
No, unless you connect Photos to a Google service that shows ads. Google's Photos privacy page says it "doesn't use your photos and videos for ads," but its privacy hub says that if you connect Photos to Search through Personal Intelligence, Search may use that data to personalize content, including ads.
Does Ask Photos keep my voice recordings?
No. Google's privacy hub says that when you ask by voice, "only the type of query (voice or text) is saved, and the audio recording won't be saved."