Meta plans to use your activity on other websites and apps to personalize Facebook, Instagram, Reels, and AI responses starting in July, and that turns ad targeting infrastructure into the machinery of the core social feed. The people most affected are not privacy absolutists. They're ordinary users who thought off-platform tracking mostly shaped ads, not the next video, post, or chatbot answer they see.
That is the trust problem at the center of this update. Meta says it isn't collecting new data, according to The Verge. But the company is expanding where existing data gets used. A signal that once helped decide which ad followed you around can now help decide what feels like organic content.
Meta's off-platform data push turns Facebook and Instagram feeds into ad targeting by another name
Meta already uses off-platform activity, including games people play and purchases they make on other websites, to serve ads. The new move expands that logic into the recommendation layer of Facebook and Instagram, including feed content and AI responses.
Here’s the hard question: if the same outside activity drives both the ad you see and the “recommended” Reel beside it, how meaningful is the distinction for users?
XOOMAR view: Meta is blurring the line between advertising systems and the product itself. That matters because users have been trained to understand ads as targeted. They may not bring the same skepticism to a video in Reels, a suggested post in Feed, or a response from Meta’s AI assistant.
Meta spokesperson Emil Vazquez told The Verge that the company previously used activity across its own apps, such as likes, views, and follows, to tailor content. That is a different category of signal. If I watch five cooking videos on Instagram, I expect more cooking videos. If I buy a tent on another site and then Instagram nudges me toward camping clips, Meta has pulled a purchase signal from outside the app into the heart of the feed.
That is not a small product tweak. It changes the perceived boundary of Facebook and Instagram.
The tent purchase example shows how Meta plans to reshape what users see next
Meta’s own example is clean because it is so ordinary. If you recently bought a tent online, you might see more camping-related Reels. No scandal. No secret file. Just a purchase becoming a content signal.
“We aren’t collecting any new data as part of this update,” Meta’s blog post says. “This is about using information that businesses already send to us to further improve your experience.”
The mechanics are simple enough. Businesses already send Meta information about user activity on their sites and apps for ad relevance. Meta now wants to use that same pool of information to personalize more of what people see across its products.
| Meta signal | Current or previous use described in sources | Expanded use under the update |
|---|---|---|
| Games played off-platform | Ad personalization | Feed, content, and AI personalization |
| Purchases on other websites | Ad personalization | Reels and other content recommendations |
| Activity across Meta apps | Content personalization through likes, views, follows | Still used, now joined by more outside signals |
| Conversations with Meta AI assistant | Used to personalize ads last year | Now sits beside broader AI personalization concerns |
The difference is psychological. Ads are labeled. Users expect them to be transactional. Feed recommendations often feel like the platform “knows” what you like, not that another business sent Meta a behavioral breadcrumb.
For builders and makers, this is also a reminder that product architecture has politics baked into it. Data pipelines are not neutral once they decide what people read, watch, buy, or ask next. We’ve seen the same principle in AI infrastructure choices, where costs and serving layers shape what gets built, as in our analysis of why your GPU bill picks the vLLM vs TGI winner.
Meta says it isn't collecting new data, but that misses the real privacy issue
Meta’s strongest technical defense is narrow: no new collection. The company says this update concerns information businesses already send. That may be true on its own terms. It also misses the point.
The privacy issue is not only collection. It is repurposing.
A user may tolerate off-platform activity being used for ads because that bargain has become visible, annoying, and familiar. That does not mean the same user gave meaningful consent for that data to tune Reels, reshape Feed, or inform AI responses. Consent buried in one context does not automatically travel cleanly into another.
What does the average user actually know about which businesses send activity to Meta, what categories Meta derives from it, and how long those signals shape what appears on screen?
The answer, based on Meta’s own control changes, is that the company expects users to manage this through settings. Meta is combining controls and will no longer offer the “Your activity off Meta technologies” setting that let people disconnect activity businesses share from their account. The relevant control becomes “Activity from other businesses”, which Meta says will cover ads and non-ads personalization.
XOOMAR view: This is where big tech privacy language often gets too clever by half. A company can make a technically accurate statement while avoiding the user’s lived experience. “No new data” sounds reassuring. “We will use data from other businesses to influence your feed and AI responses” sounds very different.
Both can describe the same update.
AI responses make Meta's off-platform tracking feel more personal and more risky
The feed is one thing. AI responses are another.
People scroll feeds with some level of suspicion. They may treat a chatbot answer as more neutral, especially when it responds in a direct, conversational format. If Meta uses outside activity to personalize those responses, users deserve to know when that is happening and what kind of signal is involved.
A camping Reel after a tent purchase is easy to understand. But what happens when an AI assistant suggests where to travel, what to buy, which creator to follow, or how to plan a weekend, and the user cannot tell whether the answer reflects a recent purchase, browsing behavior, app activity, or Meta’s inferred interests?
That hidden personalization is the risk. The system may feel like advice while behaving partly like targeting.
AI personalization needs a higher transparency standard than feed ranking because the interaction is more intimate. A feed surrounds the user with options. A chatbot often compresses options into a response. That gives the personalization layer more power.
For AI teams, this is not an abstract governance debate. The same pressure shows up when content and marketing teams decide how much automation to accept in planning and production. Our piece on AI content brief tools SEO teams will regret skipping made a related point: AI systems can improve relevance, but only if users understand the inputs and constraints behind the output.
Meta’s AI ambitions raise the stakes here. Data that once helped sell ads can now shape conversational interfaces that appear neutral. That deserves more than a buried toggle.
The strongest defense of Meta's update is convenience, and users will feel that pull
Meta has a real counterargument. Personalization often works. Many users would rather see relevant Reels, useful group suggestions, creator posts that match their interests, and ads that are at least somewhat aligned with their lives.
If someone buys camping gear, maybe they do want hiking videos. If they play certain games, maybe they do want related communities or creator clips. A less random feed can feel better.
Meta can also fairly argue that it already uses this data for ads. Extending it into content recommendations may look like a logical product improvement from inside the company. Why keep useful signals trapped in one part of the product when they could make the whole experience more relevant?
The problem is control. Convenience does not erase the need for explicit consent, clear labeling, and real opt-outs.
A useful feature becomes coercive when users must hunt through settings to avoid it. It becomes murkier when one control covers ads, feeds, and AI responses together. And it becomes harder to trust when the company frames the change around data it is not collecting rather than product surfaces it is newly influencing.
For end users, the practical takeaway is blunt: check the “Activity from other businesses” setting when the update reaches your region. Meta says the rollout is global but will exclude some locations at launch, including the European region, the UK, Brazil, Thailand, South Africa, Turkey, South Korea, Ecuador, Nigeria, and Kenya.
Meta should make off-platform personalization opt-in, not another setting users have to hunt down
Meta should require affirmative consent before using off-platform activity to personalize feeds, Reels, and AI responses. Not a passive notice. Not a renamed setting. A direct choice.
A serious privacy dashboard would show:
- Which businesses sent activity to Meta.
- What categories Meta inferred from that activity.
- Where the data is used, including ads, Feed, Reels, and AI responses.
- How to disable each use separately, without forcing an all-or-nothing choice.
That last point matters. Users may accept off-platform signals for ads but reject them for AI. They may allow them for Reels but not for Feed. Meta’s products are not one surface, and privacy controls should not pretend they are.
The market signal is plain: personalization is moving from ads into everything. Feeds, recommendations, and AI assistants are becoming different faces of the same targeting stack. Competitors do not need to say a word for this to matter. Meta is showing where large consumer platforms want to take behavioral data next.
The watch item now is whether Meta treats this as a consent problem or a settings problem. If it chooses the latter, users will get more relevance, but less clarity. If Meta wants to personalize the internet around people, it should start by treating them like people, not data exhaust.
Impact Analysis
- Meta is expanding how existing tracking data shapes the core user experience, not just ads.
- Users may not realize that off-platform behavior can influence recommended posts, Reels, and AI answers.
- The change blurs the line between advertising infrastructure and seemingly organic content discovery.
Originally published on XOOMAR. For more news and analysis, visit XOOMAR.
