Meta's AI Features Are On by Default. Here's What We Actually Enable for Health & Wellness Accounts.

In the space of a few months, Meta put an AI agent inside Ads Manager and switched on AI-powered pixel enrichment for most advertisers — the latter on an opt-out basis, with a notification window most teams missed. If you manage paid acquisition for a health or wellness brand, both of those sentences should make you sit up.

I laid out a full decision framework for this in a piece for the team at Curve: Five questions to answer before turning on Meta's new AI features. That article walks through the questions one by one — HIPAA status, patient portal exposure, condition implication, server-side infrastructure, and review capacity. This companion post is about what happens after you've answered them: the configuration defaults we actually run at Coast Studio across a portfolio of regulated accounts, and the operating rhythm that keeps those configurations honest.

Why "just turn it on" is the wrong frame

Meta's position is consistent: advertisers are responsible for making sure sensitive data doesn't reach Meta's systems. That responsibility doesn't shrink when Meta's own AI starts pulling more signal from your pages — it grows. AI pixel enrichment can extract rendered page content from wherever your pixel fires. If your site includes booking flows, intake forms, or condition-specific landing pages, "more signal" can quietly mean "sensitive information you never intended to transmit."

So we treat every AI feature as a data-flow decision first and a performance decision second. The performance upside is real in some accounts. It is never worth an uncontrolled data pathway in a category where the FTC has already shown it will act.

Our portfolio defaults, by account type

After running the five-question framework across client accounts, we landed on four standing configurations:

• HIPAA-covered entities and business associates: Both features off, full stop. Meta won't sign a BAA. Conversion signal runs through a compliant server-side intermediary with payload filtering, and nothing about that changes because a new toggle appeared.
• Consumer health & wellness (not covered, but condition-implying): AI pixel enrichment off. If buying the product implies something about a person's health — sleep aids, glucose monitors, mental wellness apps — enrichment stays off, and the agent operates in research mode only.
• Wellness-adjacent brands with clean data paths: Enrichment considered page-by-page, agent allowed for drafting and analysis, human approval required before anything ships. This only works when a server-side layer already controls what Meta receives.
• B2B healthtech and non-sensitive accounts: Broadest access — but still with a human review gate on agent-created campaigns and audiences, because staffing that review is a prerequisite, not a nice-to-have.

The two failure modes we design against

1. The silent reclassification

Meta runs its data source categorization and its AI feature eligibility as separate, opaque systems — and both can change without warning. The scenario that should worry you: you enable enrichment while classified as low-risk, and Meta later moves your domain into health and wellness. Nothing alerts you. Your account is now doing something your compliance posture assumed it couldn't. Our answer is a quarterly configuration audit on every regulated account: verify the category, verify the toggles, document the state. It takes an hour and it has caught real drift.

2. The unreviewed agent

The in-platform agent will happily build audiences and campaigns. In a regulated account, an AI-built audience is a compliance artifact — it encodes assumptions about who your customers are and what their behavior implies. If nobody on the team has capacity to review what the agent produces before it goes live, the agent shouldn't be producing things that can go live. Research-only mode exists for exactly this reason, and for most lean teams it captures most of the value at a fraction of the risk.

Where this is heading

Meta has been explicit that it wants campaign execution to become largely autonomous. Each new AI capability shifts a little more decision-making from your team to Meta's systems — which means each one deserves the same question: do we control what data this touches, and do we review what it does? Advertisers who build that discipline now will be able to adopt each new capability early and safely. Advertisers who don't will either adopt blindly or refuse wholesale, and both of those lose.

For the full pre-flight checklist — including the patient-portal and infrastructure questions I've only gestured at here — read the original piece on Curve. And if you're weighing restrictions as well as AI features, start with our companion playbook: Meta Keeps Restricting Health & Wellness Advertisers.

FAQ

Are Meta's AI features safe for health and wellness advertisers?

Not by default. Because enrichment can read page content and Meta holds you responsible for what reaches its systems, treat each feature as a data-governance decision. Covered entities should disable them; consumer health brands should default to off unless a server-side layer controls the data path.

What is AI pixel enrichment?

A default-on feature (since May 2026) that lets Meta's AI extract additional signals — including rendered page content — from pages where your pixel fires. Check whether it's active on your pixels in Events Manager's data source settings.

Is this the same as Advantage+?

No. Advantage+ automates campaign delivery. These newer features operate at the account and data-source level — they change what data Meta collects and what an AI agent can do inside your account, which is why they warrant separate review.

What should a small team do?

Keep the agent in research mode, keep enrichment off if your product implies anything about health status, and put your energy into the server-side data layer — it's the control point that makes every future Meta feature decision easier.

Can Coast Studio audit our setup?

Yes — AI feature configuration is now part of every account audit we run for regulated brands. Book a consultation and we'll map your current exposure in the first session.