Guide · 4 min read
How to Stop Social Apps Using Your Posts to Train AI
Turn off the AI-training switches that exist, file objections where a form is offered, and make old public posts private, because opt-outs only cover the future.
Short answer
Turn off the AI-training setting on each app that has one (X and LinkedIn do), use Meta's objection form if it's offered where you live, and make public posts private or delete them, because no opt-out removes data that was already used.
Key takeaways
- X and LinkedIn both have a settings switch that stops your future posts and activity being used to train their AI models.
- Meta's apps offer an objection form in some regions, mainly the EU and UK; elsewhere the practical option is limiting what's public.
- Opting out works going forward only. Content already used for training stays in models that were trained on it.
- Private accounts, deleted posts and fewer public photos reduce what any company or scraper can collect.
Most big social apps now use posts, photos and activity to build their AI products. Some let you say no, some only let you object in certain countries, and none can take back what was already used. The practical plan is short: flip the switches that exist, file the forms where they're offered, and cut down what you leave public.
Which apps use your posts for AI training?
Most major social apps say in their privacy policies that public posts and activity can be used to develop their AI models.
The wording differs, but the pattern is similar. X uses public posts and interactions to train Grok. LinkedIn uses profile data and public posts for its generative AI features in many regions. Meta uses public posts, photos and comments from Facebook and Instagram, plus conversations with Meta AI. Policies change often, so treat the specifics below as a snapshot as of November 2026 and check each app's privacy center before you rely on them.
| App | Is there a settings switch? | Where to look |
|---|---|---|
| X | Yes | Settings and privacy, Privacy and safety, Grok (data sharing) |
| Yes | Settings, Data privacy, Data for Generative AI Improvement | |
| Facebook and Instagram | Objection form in some regions | Meta Privacy Center, generative AI section |
| Threads | Follows Meta's policy | Meta Privacy Center |
| Bluesky, Mastodon | No company model training described | Server rules and posts stay public to scrapers |
How do you turn it off on X?
On X, open Settings and privacy, then Privacy and safety, and turn off the option that lets your posts and Grok interactions be used for training.
The option sits in a Grok or data sharing section. If it's missing on the app, try the web version at x.com, where settings sometimes appear first. Turning it off covers future use. While you're there, check whether your posts are protected; a protected account limits who can read your posts in the first place.
How do you turn it off on LinkedIn?
On LinkedIn, go to Settings, Data privacy, and switch off Data for Generative AI Improvement.
LinkedIn turned this on by default for members in many regions. Switching it off doesn't change how recruiters or contacts see your profile. If you're in a region where LinkedIn offers it, there's also a data processing objection form for data collected before you switched the setting off.
What about Facebook, Instagram and Threads?
Meta doesn't offer a simple on/off switch everywhere; in the EU and UK it has offered an objection form, and elsewhere your main lever is keeping posts non-public.
Look for the generative AI section of the Meta Privacy Center. If an objection form is available to you, submit it once for your accounts; Meta has said earlier objections carry over. If no form is offered where you live:
- Set your Instagram account to private.
- On Facebook, change the audience of future posts to Friends, and use the "Limit past posts" tool to change old public posts in one go.
- Avoid typing personal details into Meta AI chats, which Meta treats differently from your posts.
What can an opt-out not do?
An opt-out stops future use; it doesn't pull your posts back out of a model that was already trained on them.
That's the uncomfortable part. Old public posts may already be in a training set, and copies may sit with outside scrapers that never asked any app's permission. The only thing that shrinks future exposure is shrinking what's public. For many people that means:
- Deleting or archiving old posts you no longer need public.
- Removing photos of your face, your home or your kids from public albums.
- Keeping a separate, deliberately public account for work, and a private one for everything else.
The privacy settings checker walks through the related switches on each app, including who can find you, connected apps and ad personalization. Our guide on reducing what social apps know about you covers the rest of the data trail.
Do decentralized networks train AI on posts?
Bluesky and Mastodon don't describe training their own models on your posts, but public posts there can still be collected by anyone.
Public posts on open networks are readable by design. Bluesky has published a proposal for letting users signal whether they consent to AI training, and some Mastodon servers block known AI crawlers. Those are signals, not locks. If a post must not end up in a dataset, the safest version of it is one that isn't public.
How often should you check these settings?
Check them every few months and after any privacy policy update, because AI settings are among the most frequently changed.
When an app emails you about a new privacy policy, that's usually the moment a new AI use starts, and sometimes when a new opt-out appears. Five minutes then saves a lot of guessing later.
FAQ
If I opt out, will my old posts be removed from the AI model?
No. Opt-out settings stop future use. Data that was already used to train a model stays part of that model, and companies generally don't retrain to remove one person's posts. LinkedIn offers a separate objection form for some regions, but it doesn't promise to undo past training.
Does making my account private stop AI training?
It helps a lot. Most AI-training policies on social apps describe public content. Private or protected posts are also harder for outside scrapers to collect. Check each app's privacy policy, because some include your activity and interactions with their AI assistant too.
Do messages get used to train AI?
End-to-end encrypted messages can't be read by the company, so they can't be used for training. Unencrypted DMs and chats with an app's built-in AI assistant are a different matter, and policies vary, so treat anything you type to an AI chatbot inside a social app as something the company can read.
Why is the opt-out switch greyed out or missing?
Settings roll out by country and account type, and apps move them often. Search the app's settings for words like "AI", "Grok", "generative" or "data sharing". If nothing turns up, the app's privacy center usually explains what applies in your region.