Meta AI vs Gemini privacy: training, ads, reviewers, and deleting, with Morse alongside
The Morse teamPublished Facts checked 14 min read
On Meta AI vs Gemini, Gemini is the more private of the two as of November 2026, but neither is private from the company that runs it. Google's help pages now say Gemini's training settings "are off by default" and that your chats "are not being used to show you ads," while Meta says it uses your interactions with its AIs to improve its AI, names no training switch for Meta AI in the US on the pages we read, and has used AI chats to pick ads on Facebook and Instagram since December 16, 2025. Gemini falls behind on how long it keeps things. A chat Google's reviewers pick is kept up to three years, and even with History off, Google holds each chat with your account for around two months. Meta's one clear win is Incognito Chat, which Meta says it "can't access." Choose Gemini if you want training and ads off without touching a setting, Meta AI if you'll keep sensitive questions in Incognito Chat, and Morse if you want Gemini without the provider knowing it's you.
We read both companies' own pages on November 18, 2026. From Google: the Gemini Apps Privacy Notice (last updated November 16, 2026), the Privacy Hub's questions (updated November 17), its help pages on history, memory, temporary chats, sharing, and connected apps, the Google AI plans page, and an Internet Archive copy of the hub's October 26 version. From Meta: its AI terms (effective May 13, 2026), the privacy policy's current version (effective July 23, 2026), its December 2025 ads post, four Meta AI help pages, a Privacy Center page on saved details, a WhatsApp help page, and the launch post and two help pages for Muse, its new agent. Two Facebook Privacy Center pages wouldn't load for us, so nothing below rests on them. We ran no tests and looked only at personal accounts. Both companies earn credit. Google puts a number on nearly every copy it keeps and has now turned training off by default, and Meta built a private mode that it says keeps even Meta out.
Meta AI vs Gemini vs Morse, side by side
As of November 18, 2026, for a personal account in the US. Each Meta AI cell links to Meta's or WhatsApp's own page and each Gemini cell to Google's. The Morse column is drawn from our privacy page.
| Question | Meta AI | Gemini | Morse |
|---|---|---|---|
| Where your prompt is read | Meta receives your prompts (WhatsApp); Incognito Chat runs in Private Processing (Meta) | On Google's servers (Google) | Anonymous: at the provider, without your identity. Private: at a host with a zero data retention agreement. E2EE: only inside a verified enclave |
| Who checks that | For Incognito Chat, Meta offers a Private Processing report made on your device (Meta) | No outside check described on the pages we read | Your browser checks the enclave's attestation before an E2EE message is sent |
| Where history lives | In the History section of your Meta AI account (Meta) | In your Google Account, auto-deleted after 18 months by default (Google) | In your browser, encrypted, not on Morse's servers |
| Training on your chats | Yes: "Meta uses your interactions with AIs to improve AI at Meta" (Meta) | Off by default since Google's November 2026 update (Google) | Morse doesn't train AI on your chats |
| A switch for training | None on the Meta AI pages we read; Muse has one, on at first use (Meta) | Chat content including text & uploads, under Settings, then History (Google) | None needed |
| Chats used for ads | On Facebook and Instagram since December 16, 2025, in "most regions" (Meta) | No, except Google shopping cart data saved to Wallet (Google) | No ads on any plan |
| Human review | "Automated or manual (human)," sometimes through third-party vendors (Meta) | A subset of chats, kept up to three years (Google) | Morse's servers don't keep messages; in Anonymous mode the provider may keep a message for a limited time to check for abuse, under its own policy |
| After you delete | "May not delete our copy" (Meta) | Removed from view, then storage; reviewed chats stay (Google) | Clearing your browser or deleting your account deletes your chats, and Morse can't recover them |
| Most private mode | Incognito Chat: "Meta can't access it" (Meta) | Temporary chat, kept around two months (Google) | E2EE on Pro; Temporary chats kept in memory only |
| Models | Muse Spark, which Meta says "now powers the Meta AI assistant in the Meta AI app and meta.ai" and is coming to Meta AI in WhatsApp, Instagram, Facebook, Messenger, and Threads (Meta) | Google's Gemini models (Google) | 80+, open and closed, including GPT, Claude, and Gemini |
| Price | Free; Meta One Core $7.99 and Premium $19.99 a month add more AI media (Meta) | Free; AI Plus $4.99, AI Pro $19.99, AI Ultra from $99.99 a month (Google) | Free, 15 messages a day; Pro $19 and Max $49 a month |
Is Meta AI or Gemini more private?
Gemini, in three of the four places most people worry about. Google starts with training off, keeps chats out of ads, and writes down how long each copy lasts. Meta trains on Meta AI chats, uses them to pick ads, and its AI terms warn that "deleting individual messages, entire threads, or even your account may not delete our copy of your personal information" (Meta).
Meta AI wins on the fourth, with a mode that keeps the company out. Meta's help page says of Incognito Chat, "Only you can see the chat. Meta can't access it," and that conversations "won't be saved, stored, or appear in your Meta AI chat history" once you exit (Meta). Gemini's temporary chat is a weaker promise. It stays out of your history and training, but Google keeps it with your account "for around two months" and uses it "to protect Google, our users, and the public," including with human reviewers (Google). Incognito Chat still sends search queries to "select search engines without connecting the search to your account," and web search stays on until you toggle it off (Meta).
Do Meta AI and Gemini train on your chats?
Meta AI does, and Gemini doesn't unless you opt in. Meta's help pages say it uses your AI interactions "to improve AI at Meta" (Meta), and WhatsApp's page says Meta's models are trained on data including "the interactions and prompts you send to AIs from Meta" (WhatsApp). None of the Meta AI pages we read offers a US user a switch. Meta's AI terms point people in the UK, Brazil, and the European Region to their own versions (Meta), and our page on whether Meta AI is private covers the regional rules.
Meta built that switch for its own new agent. Muse, launched September 8, 2026, runs in its own app or "directly in WhatsApp" (Meta). For Muse, Meta says "This setting is on when you first use Muse," that you turn it off under Settings, Data controls, then Help improve our AI models, and that "Changes to this setting also apply to previous interactions" (Meta). Meta's Muse pages describe that switch for Muse alone, and we found no matching one for Meta AI in Facebook, Instagram, or WhatsApp.
Google's history page now says its two training boxes, one for audio and Gemini Live and one called Chat content including text & uploads, "are off by default," and that you need History on to turn them on (Google). Two exceptions stay, both from the Privacy Hub: feedback still trains models, and "Your Gemini settings don't control processing of your chats to create anonymized data to improve Google services" (Google).
Some privacy rankings haven't caught up. Incogni's ranking, still online under a "2025 [2026]" title, says "Gemini, DeepSeek, Pi AI, and Meta AI don't seem to allow users to opt out of having prompts used to train the models" (Incogni). As of November 18, 2026, that's still true of Meta AI on Meta's US pages, but Gemini's training settings start off.
What did Google's November 2026 update change?
It turned training off by default, and it stretched the copy Google keeps with History off from 72 hours in the hub's October 26 version (Internet Archive copy of Google's hub) to "around two months" in its current version (Google). Every earlier version of Google's Gemini rules is in our Gemini privacy policy history.
| Date | Company | What changed | Source |
|---|---|---|---|
| December 16, 2025 | Meta | AI chats start shaping content and ads on Facebook and Instagram | Meta |
| May 13, 2026 | Meta | Current AI terms take effect | Meta |
| July 23, 2026 | Meta | Current privacy policy takes effect | Meta |
| September 8, 2026 | Meta | Muse launches, with its own training switch and no ad-system sharing | Meta |
| September 15, 2026 | Meta | Meta One plans launch; "Meta AI will still be free for everyday use" | Meta |
| October 26, 2026 | Hub says Keep Activity-off chats are kept 72 hours | Internet Archive | |
| November 16, 2026 | Gemini Apps Privacy Notice updated, then the hub: the setting is now Gemini History, History-off chats are kept around two months, and the history page says training starts off | Google, Google |
Google's pages don't all use the new name yet. Its memory page still says you need "Keep Activity" on (Google), and its connected apps page still says app data trains models "if your Keep Activity setting is on" (Google). Check your own boxes under Settings, then History.
Do Meta AI and Gemini use your chats for ads?
Meta AI does, and Gemini doesn't. Meta's post gives its own example: chat with Meta AI about hiking and "you might start seeing recommendations for hiking groups, posts from friends about trails, or ads for hiking boots" (Meta). It covers "a voice chat or a text exchange." Meta says it doesn't use chats about religion, sexuality, politics, health, ethnicity, philosophy, or union membership to show you ads, and that Meta AI chats on a WhatsApp account you haven't added to the same Accounts Center aren't used "to personalize experiences across different accounts." Its post points to Ads Preferences and feed controls to adjust what you see, and names no setting that keeps your chats out. Muse is the exception inside Meta: "Muse doesn't share your conversations or the data in your virtual machine with Meta ad systems" (Meta).
Google's line is "Your Gemini Apps chats are not being used to show you ads. If this changes, we will clearly communicate it to you" (Google). Shopping is the exception. Cart and order data from Gemini goes to your Google Wallet, where Google may use it "for personalization and ads (depending on your settings)."
Do people at Meta and Google read your chats?
Some of them, at both. Meta's AI terms say "In some cases, Meta will review your interactions with AIs," by automated or human review, sometimes "through third-party vendors" (Meta). Meta adds that when its AIs can't answer, it shares "the messages you send to AI and general information, like your region, with select partners" (Meta).
Google is more exact about how long. Reviewers, "including Google's trained service providers," read a subset of chats, which are "disconnected from your account" first and then kept up to three years, along with "your language, device type, location info, or feedback" (Google). Turning training off doesn't end review: "Regardless of your settings, Google uses your chats to respond to you and help protect Google, our users, and the public, including with help from human reviewers." The companies doing that work are in who reviews AI chats.
What happens when you delete a Meta AI or Gemini chat?
You lose your view of it, and both companies may keep a copy. In the Meta AI app, tap Menu, Settings, Data & privacy, Manage your information, then Delete all chats and media (Meta). That clears your History and Media, but "Your public vibes won't be deleted," and Meta's terms still say your deletion may not reach its copy. We cover each Meta app's steps in how to delete Meta AI chats.
Google says it first removes a deleted chat from view, then starts deleting it from storage (Google). Reviewed chats are "not deleted when you delete your history." A shared chat can outlast your delete too. Anyone with the link can continue most shared chats, and deleting the link "doesn't delete the chat from other's Gemini history" (Google). Data from Gemini's tasks, in its remote browser and remote computer, survives a history delete until you clear it in Task settings (Google). Step-by-step paths are in how to delete Gemini history.
What do Meta AI and Gemini remember about you?
Meta AI starts from your Meta profile, and Gemini starts from your Google history. Meta says your "age and gender, and interests based on your activity across our products," including "the types of content, including ads, you view or interact with," personalize your AI chats, along with your location (Meta). Its Privacy Center adds that an AI "can remember details about you and context from the conversation," such as liking Italian food (Meta). On Muse, Meta warns that "After you delete something from Muse, Muse may still remember information it learned from what you deleted" (Meta).
Gemini's Memory draws on chats in your history, and "Gemini may use sensitive info you provided in your chats" when it's on (Google). You can see and edit a memory summary, or ask, "What do you know about me?" To check whether a reply drew on old chats, Google suggests asking, "Did you use any info from past chats?" (Google). Task memory, for eligible users, keeps "your workflows, project context, preferences, people in your world, and personal details," and deleting a task memory may leave the action that created it in your history (Google).
What can Meta AI and Gemini connect to?
Gemini reaches further into your accounts. With Personal Intelligence, you can connect Google Workspace (Gmail, Calendar, Drive, and more), Search services, Contacts, Google Wallet, Google Photos, and YouTube (Google). Google says it uses that data to "improve Google services, including training generative AI models, for everyone," and that Photos data is used "to infer your interests, relationships to people in your photos, and where you've been." With History off, only Device assistance, Gemini Notebook, Messages, Phone, and WhatsApp keep working (Google), and the hub names WhatsApp, Phone, Messages, and Device assistance as the ones on Android (Google). Cutting an app off later leaves what Gemini already has, since disconnecting it "doesn't delete any data from Gemini Apps Activity."
Meta AI lives inside apps you already use. WhatsApp says "Only your interactions and prompts sent to AIs from Meta or that people choose to share with AI's from Meta, can be read by Meta," and that Meta also receives your device, account, usage, and general location information (WhatsApp). On the pages we read, Meta's email and calendar connectors belong to Muse, which "won't download your whole inbox, unless you ask it to" (Meta).
Does paying for Meta AI or Gemini make it more private?
No, on the pages we read. Google AI Plus, Pro, and Ultra raise your usage limits (Google), and the Gemini notice treats your subscription as one more piece of "subscription related information" it collects (Google). Meta One's Core plan lets you "generate more images and videos with Meta AI," and Meta says "Meta AI will still be free for everyday use" (Meta). None of the pages we read gives paying customers different training, review, or ad rules.
How Morse keeps your questions out of Meta's and Google's accounts
Most rows above describe a copy that Meta or Google holds, filed under an account that already knows you: for Meta, the identity its Facebook and Instagram ads are sold against; for Google, the account that also has your Gmail, Google Photos, and search history. Settings only decide what each company does with that copy. Morse's servers don't keep your messages or log their content.
Morse doesn't train AI on your chats, shows no ads on any plan, and runs no advertising trackers or third-party analytics.
- Anonymous, to ask Gemini without the provider knowing it's you. Gemini is one of the proprietary models in Anonymous mode, alongside GPT and Claude. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy. Your words still reach the provider, so leave out names and details that point back to you.
- E2EE, for the questions you'd otherwise save for Incognito Chat. Your browser checks the secure enclave's hardware attestation before anything leaves your device, then encrypts the message so only that enclave can open it, and Morse's servers relay text they can't read. The enclaves are run by third-party providers, not by Morse, and the protection rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models, with no web search or image generation, and comes with Pro. Here's how end-to-end encrypted AI chat works.
- Private, the default, for everyday questions you'd rather no host held onto. Open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok, run here. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
Your chat history is stored in your browser, encrypted, not on Morse's servers, and it locks when you sign out. Morse doesn't sell your data. A Temporary chat is kept in memory only. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or losing your device deletes your chats, and Morse can't recover them, so save anything you'll need again. The privacy page shows who can see what in each mode.
Choose Meta AI, Gemini, or Morse
- Choose Meta AI if you already live in WhatsApp and Instagram, you'll take sensitive questions to Incognito Chat, and you accept that your other AI chats shape your ads.
- Choose Gemini if you want training off and ads out of your chats from the start, and Gemini working across Gmail, Drive, and Photos, knowing reviewed chats last up to three years.
- Choose Morse if you want Gemini, GPT, and Claude without the provider knowing it's you, E2EE your own browser checks, Morse not training AI on your chats, no ads on any plan, and history kept encrypted on your device.
Start with a guest chat on the Morse home page, or make a free account for 15 messages a day with no card; Pro adds your choice of model and E2EE for $19 a month.
Morse publishes this comparison. We've tried to keep it fair; if something's wrong or out of date, email hello@trymorse.ai.
Meta AI is a trademark of Meta Platforms, Inc. Morse is not affiliated with or endorsed by Meta Platforms, Inc.
Gemini is a trademark of Google LLC. Morse is not affiliated with or endorsed by Google LLC.
Common questions
Is Muse the same thing as Meta AI?
No. Meta launched Muse on September 8, 2026 as a separate personal AI agent, in its own app and inside WhatsApp, and its help pages set different rules: a training switch you can turn off under Data controls, and no sharing of Muse conversations with Meta's ad systems.
Can Meta AI or Gemini read my WhatsApp messages?
Only what you send them. WhatsApp says Meta can read only your messages to its AIs and what you choose to share with them. Gemini can use WhatsApp as a connected app on Android, and Google lists it among the few apps that keep working with Gemini History off.
Does paying for Meta AI or Gemini make it more private?
Not on the pages we read. Google's paid plans raise usage limits, and the Gemini privacy notice lists your subscription as one more thing it collects. Meta One's Core plan adds more Meta AI image and video generation, and Meta says Meta AI stays free for everyday use.
Can I ask Gemini what it knows about me?
Yes. Google says you can ask Gemini 'What do you know about me?' to see its memory summary, and 'Did you use any info from past chats?' to check whether a reply drew on your history.