How to Use Manus AI in Meta Ads Manager: A Practitioner’s 2026 Setup Guide (And What the China Block Changes)

If you opened Ads Manager this morning expecting Manus AI to look the same as it did yesterday, today’s Bloomberg headline just rewrote the question. China’s NDRC ordered Meta to unwind its $2 billion Manus AI acquisition (Bloomberg, April 27, 2026). The integration that launched on February 17 is still live in the platform — for now. But the answer to “should I connect it?” just changed shape.

I’ve been using Manus inside Ads Manager since the integration shipped, and the surprising thing is how little of the workflow actually depends on the corporate ownership question. The connection steps haven’t changed. The five prompts that save my Monday morning reporting block haven’t changed. What’s changed is the timeline question: how long do you have, and what should your fallback look like?

This is the practitioner setup guide nobody published before today’s news, plus the decision framework you actually need now.

TL;DR: Manus AI launched inside Meta Ads Manager on February 17, 2026 as a read-only intelligence layer (MediaPost, 2026). Meta acquired Manus for over $2 billion in December 2025 (CNBC, 2025), but China’s NDRC ordered the deal unwound on April 27, 2026 (Bloomberg, 2026). The integration still works today. Connect it via Settings → Integrations → Manus AI, run the five reporting prompts below, and don’t yet rebuild your stack around it.

What Is Manus AI Inside Meta Ads Manager?

Manus AI is a read-only intelligence layer that connects to your ad accounts and answers questions about performance using natural language ([Search Engine Land](https://searchengineland.com/meta-adds-manus-ai-tools-into-ads-manager-469410), 2026). Meta describes it as an “intelligence and workflow layer” — the panel sits inside Ads Manager but does not change bids, pause ads, move budget, or create anything ([AdAmigo](https://www.adamigo.ai/blog/manus-ai-meta-ads-manager-capabilities-limitations), 2026). It reads, summarizes, and explains. That’s the entire surface area.

The underlying agent comes from Butterfly Effect, the Wuhan-based parent of the Monica AI assistant, which launched the original Manus agent in March 2025. Inside Meta’s product, Manus runs Claude 3.5 Sonnet plus fine-tuned Qwen models behind the scenes. Outside Meta, it’s a general-purpose web agent on a freemium plan (Manus pricing — free tier gives 300 daily credits; Pro starts at $20/month).

The version inside Ads Manager is gated to your connected ad accounts. There is no separate subscription. If you log in, you have it.

What surprised me on first use was how little it tries to do. There’s no autonomous loop, no “run this every Monday” scheduler. You ask, it answers, you act. Jon Loomer called the integration “not a direct integration” and “confusing and sloppy” in his February 18 review (Jon Loomer, 2026) — and he’s not wrong about the UX. But he undersold the leverage on reporting tasks, which is where Manus earns its keep.

Why Did China Block Meta’s Manus Acquisition Today?

China’s National Development and Reform Commission ordered Meta to divest Manus on April 27, 2026, citing national security and data sovereignty concerns over a U.S. company controlling a Chinese-origin AI agent with deep enterprise access (Bloomberg, 2026). The deal closed in late December 2025 (TechCrunch, 2025), but the regulatory review continued, and the unwind order makes the Manus-inside-Ads-Manager integration a question of months, not years.

What happens to the Ads Manager panel? Meta hasn’t said. Three plausible paths:

1. Status quo through the appeal window. Meta has signaled it will challenge the order. The integration likely runs unchanged through Q3 2026.
2. Engineering rebuild on Meta’s own agent stack. Meta’s Ranking Engineer Agent shipped to engineering.fb.com in March 2026 — the in-house agent muscle exists. The Manus panel could be quietly rewritten on top of REA-class infrastructure within two quarters.
3. Sunset and replacement. The panel could be deprecated entirely, with the AI Business Assistant ([Marketing Brew](https://www.marketingbrew.com/), April 24, 2026 GA) absorbing the conversational reporting use case.

The point: build a workflow that survives any of these three outcomes. Don’t build your reporting cadence around a UI panel that might not exist in nine months.

How Do You Connect Manus AI to Meta Ads Manager Step-by-Step?

The connection lives at manus.im/dashboard/integrations if you’re starting from the Manus side, or under Settings → Business Integrations → Manus AI if you’re starting from Ads Manager. Both paths land in the same OAuth screen. The connection is read-only — Meta enforces this at the permission layer, not as a Manus policy (ALM Corp, 2026).

Here are the steps, in order:

1. Open Meta Ads Manager as a Business Manager admin. The integration requires admin role on the ad account; standard advertiser access won’t see the panel.
2. Click the Manus icon in the right-side toolbar. If you don’t see it, go to Settings → Business Integrations and toggle “Manus AI” on for the relevant ad accounts.
3. Authorize the connection on the OAuth screen. Manus requests ads_read and ads_management scopes — but Meta’s product layer disables write actions even when the scope is granted. You can verify this by trying to ask Manus to “pause this ad” — it will refuse.
4. Select ad accounts to expose. If you manage multiple clients, connect them one at a time. Manus does not isolate prompts across accounts within a single chat — context bleed is real.
5. Set your default time window. The default is “last 7 days.” For DTC brands with weekly creative refreshes, change this to “last 14 days” so winner identification has enough sample.
6. Test with a baseline question. Ask: “What were my top three ad sets by ROAS in the last 14 days, and what creative themes did they share?” If Manus answers with named ad set IDs and a coherent creative summary, the connection is healthy.

The whole setup takes under 10 minutes per account. The expensive part isn’t the connection — it’s training yourself to ask the right questions, which is the next section.

What Are the 5 Manus Prompts That Actually Save Reporting Time?

Gartner forecasts that 40% of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5% in 2025 (Gartner press release, August 2025). Most of those agents will get used for the wrong tasks. The pattern that actually pays back the time investment is narrow, repeatable reporting questions — not open-ended strategy chats.

These are the five prompts that earn their keep. Each one replaces a manual exploration step I used to do in Ads Manager filters or Looker.

1. “Show me every ad set with ROAS below [threshold] AND spend above [floor] in the last 14 days, ranked by total wasted spend.” This is the kill list. Manus returns a sortable summary you can act on in Ads Manager directly. The “wasted spend” framing focuses you on dollars at risk, not just bad ratios.
2. “Compare creative theme performance across the last 14 days vs the prior 14 days. Group by hook style, format, and CTA.” Manus has to interpret theme — if your naming convention is sloppy, the answer will be sloppy. This forces you to actually use ad name conventions, which is the secondary benefit.
3. “Which audience segments saw the largest CPA delta this week, and were any of them inside Advantage+ exclusion sets?” The exclusion-set check is the part that surprised me. Manus reads the audience config, not just the performance numbers. It catches drift between what you intended to exclude and what’s actually live.
4. “What learning-phase ads exited learning this week, and what was their performance trajectory in the first 48 hours post-exit?” This is the early-winner detection prompt. Manus surfaces ads that just left learning so you can scale them while they still have momentum.
5. “Summarize this week’s top three account-level risks based on spend velocity, frequency, and CPM trends. Format as a one-paragraph Monday morning brief.” The format instruction matters. Without it, Manus returns a wall of bullets. With it, you get a brief you can paste into Slack or a client report in 30 seconds.

None of these prompts ask Manus to do anything. They ask it to summarize state. That’s the lane where it works. The moment you ask “and what should I change?” the answers get generic — Manus inside Ads Manager isn’t optimizing for your business goals, it’s pattern-matching against published Meta best practices.

What Can Manus AI Not Do in Meta Ads Manager?

Manus AI cannot change bids, pause ads, move budget, create campaigns, edit creative, or modify audiences inside Meta Ads Manager — the integration is strictly read-only (AdAmigo, 2026). Jon Loomer’s February 18 testing confirmed this: every write action returns a friendly refusal explaining that “execution is reserved for advertiser confirmation” (Jon Loomer, 2026).

This sounds like a limitation. It’s actually a feature for anyone who’s run accounts at scale. The downside risk of an autonomous agent making bid changes inside a $50K/day budget is much larger than the upside of saving a click. Read-only is the right starting position, and Meta knew it.

Other things Manus can’t do that surprised me:

• Cross-account analysis. Each chat is scoped to one ad account. If you manage a portfolio, you can’t ask “which client account has the worst CPA trend this month” inside a single prompt.
• Custom attribution. Manus reads Meta’s default attribution settings. If your attribution lives in a separate analytics layer, the answers won’t reflect it.
• Historical comparisons beyond 90 days. The data window respects Meta’s standard reporting limits.
• Creative-level Andromeda signals. Manus can describe creative performance, but it does not surface the embedding-space proximity that Meta’s Andromeda algorithm uses for retrieval. For that depth, see how Andromeda reads creative.

The mental model that helps: Manus is a faster way to ask the questions you’d ask in Ads Manager filters anyway. It is not a replacement for the analyst layer above it.

Should You Still Connect Manus AI Now That China Blocked the Deal?

Yes, with one caveat: don’t rebuild your reporting stack around it. Connect it for the time-savings on weekly summaries, but keep your existing reporting infrastructure intact and document your prompts so they’re portable. Gartner forecasts that over 40% of agentic AI projects will fail by the end of 2027, often due to inadequate planning around governance and platform risk (Gartner, January 2026). Today’s NDRC order is exactly the kind of platform risk that hits the projects that didn’t plan.

The decision tree I’d use right now:

• Connect Manus and use the five reporting prompts above. The leverage is real. The integration works today. Worst case it gets sunsetted in nine months and you lose 30 minutes a week.
• Do NOT migrate your client reporting deliverables onto Manus exports. If a client weekly comes from Manus and Manus disappears, your weekly is broken.
• Do NOT pay for Manus Pro outside the Ads Manager integration. The integration is free; if Meta sunsets it, the standalone Manus product still exists, but the value prop is different.
• Document your prompt library. The prompts are portable. The exact words you use in Manus will mostly work in Meta’s AI Business Assistant, in ChatGPT with a Meta Ads connector, or in a custom agent. The prompts are the asset, not the panel.

One frame I keep coming back to: advertiser control shrinks as you move down Meta’s stack. The Ads Manager UI is the last three millimeters of a system that’s a hundred miles tall. Manus is a faster way to look at those three millimeters. It is not a way to control the rest. Build accordingly.

How Does Manus AI Fit Into Your Existing Reporting Cadence?

The integration earns its keep when you bolt it onto an existing weekly cadence, not when you replace one. Marketing Brew‘s coverage of the broader Meta AI rollout reported a 12% reduction in cost-per-result after applying the AI Business Assistant’s opportunity-score recommendations across early-access advertisers, following its April 24 general-availability launch (Marketing Brew, 2026). The Manus reporting layer sits one step earlier in the workflow — it tells you what’s happening before the assistant tells you what to change.

Here’s how I’d slot it into a Monday morning cadence that already works:

1. 9:00 AM — Open Ads Manager and the Manus panel side by side.
2. 9:05 — Run the Monday brief prompt (#5 from above) per ad account. Paste output into your reporting doc.
3. 9:15 — Run the kill list prompt (#1) and screenshot the table. Decide pause/keep before opening anything.
4. 9:25 — Run the early winner prompt (#4). Mark scaling decisions on the screenshot.
5. 9:35 — Execute decisions in Ads Manager directly. Manus can’t do this for you — that’s the Loomer point — but the prep work is done.

The total time saved versus filtering Ads Manager manually: roughly 20-30 minutes per account per week, in my use. The compounding value is consistency — the brief format is the same every Monday, which means historical comparisons get easier over time.

For deeper reporting infrastructure that doesn’t depend on any single AI panel, see my guide to AI-powered Meta ads reporting dashboards and the broader framework for AI agents inside Ads Manager. Both predate the China block and remain unchanged by it.

What Does the China Block Mean for Other Meta AI Tools?

The block is specific to the Manus acquisition. Meta’s own AI surfaces — Advantage+, the Ranking Engineer Agent, the AI Business Assistant — are unaffected. Gartner forecasts that 60% of brands will use agentic AI to deliver streamlined one-to-one interactions by 2028 (Gartner, January 2026). Meta’s own agent stack is the one most likely to deliver that for advertisers, because Meta owns the data layer.

The Manus episode is a reminder that platform risk in AI tooling is real and concentrated. The tools you control (your prompt library, your reporting templates, your data warehouse) are durable. The tools that live inside someone else’s product (the Manus panel, an agent feature ship) are not. Plan your stack accordingly. For the full picture of where each surface fits, see my mapping of the Meta AI agent stack in 2026.

Frequently Asked Questions

Is Manus AI free inside Meta Ads Manager?

Yes. The integration that launched February 17, 2026 is free to all advertisers with admin access to a Meta ad account (MediaPost, 2026). The standalone Manus product at manus.im has its own freemium pricing — free tier with 300 daily credits, Pro at $20/month with 4,000+ credits (Manus pricing) — but those tiers do not apply inside the Ads Manager surface.

Can Manus AI pause ads or change bids in Meta Ads Manager?

No. The integration is strictly read-only (AdAmigo, 2026). Manus can analyze, summarize, and explain account performance, but it cannot pause ads, change bids, move budget, create campaigns, or edit creative. Meta enforces this at the product layer regardless of OAuth scope. Jon Loomer’s testing confirmed this in February 2026 (Jon Loomer, 2026).

What happens to Manus AI in Ads Manager after the China block?

Meta hasn’t published a roadmap response to the April 27 NDRC order (Bloomberg, 2026). The integration remains live as of publication. Three likely paths: status quo through the appeal window, an engineering rebuild on Meta’s in-house agent infrastructure, or sunset in favor of the AI Business Assistant. Plan for any of the three.

How is Manus AI different from Meta’s AI Business Assistant?

The AI Business Assistant launched in general availability for all advertisers on April 24, 2026 with a published 12% cost-per-result reduction across early-access accounts (Marketing Brew, 2026). It makes recommendations and suggests changes. Manus reads, summarizes, and explains. The Assistant is the “what should I change” layer; Manus is the “what’s happening” layer. They sit at different points in the workflow.

Should I disconnect Manus AI from my Meta ad accounts now?

No. The integration is read-only and presents no incremental data risk versus other Business Manager integrations. Disconnecting it would only cost you the time savings while the panel remains operational. The decision to abandon Manus is a product-roadmap decision Meta will make, not one you need to preempt.

For a deeper dive, see my guide on meta’s ai business assistant just rolled out to every advertiser — here’s what it actually does (and what it can’t).

For a deeper dive, see my guide on meta ads ai connectors: how the one-click data layer changes reporting and attribution workflows.

The Bottom Line

Manus AI inside Meta Ads Manager is a fast, narrow, read-only intelligence layer that earned its place in my Monday morning workflow before today’s headline and remains useful after it. The China block introduces real platform risk — but the right response is portability, not avoidance. Connect the integration. Use the five prompts. Document your prompt library so it survives any panel change. And don’t build a client reporting deliverable on a foundation that might not exist in nine months.

The broader lesson is the one that keeps showing up: in a stack you don’t own, the durable assets are the ones you can rebuild somewhere else. Your prompts, your templates, your decision frameworks. Not the panel.

• Connect Manus today. The leverage is real and the cost is zero.
• Use the read-only lane. Don’t ask it to act; ask it to summarize.
• Build portable prompts. They’ll move to the next AI surface when this one shifts.
• Don’t replace your reporting stack. Augment it.

For more on building an AI-native Meta ads workflow that survives platform shifts, start with my full playbook for using AI in Meta ads and the 2026 DTC strategy framework.