Manus AI testing shows mixed results in Meta Ads Manager

Meta began rolling out Manus AI inside Meta Ads Manager last week, outlining where advertisers could find it under the Tools menu and describing its stated capabilities.

In the weeks since the rollout began, advertisers who gained access inside live accounts have started testing it beyond surface-level exploration. Availability still appears gradual. As the integration moves from announcement to application, the conversation has shifted. The theoretical promise of an AI agent integrated inside Ads Manager is now being measured against what it actually delivers when applied to real campaign data.

A detailed breakdown by Cody Schneider of Graphed moves the discussion far beyond Meta’s initial positioning. The publication reviewed feedback and experiments conducted by advertisers and AI testers. The analysis aggregated documented use cases, performance breakdowns, and user reports to evaluate where Manus performs well and where it falls short.

Where Manus stands out

Faster reporting and workflow efficiency

One of the clearest advantages identified in testing is speed. In traditional Meta Ads Manager workflows, pulling insights often requires navigating dashboards, adjusting filters, exporting reports, and manually interpreting results. That process can be time consuming, especially for advertisers managing multiple campaigns.

Testers found that Manus compresses much of that work into a single interaction. Instead of moving through multiple reporting views, advertisers can ask for a summary in plain language, and the system generates a structured response inside the interface.

For example, an advertiser can type a request such as, “Summarize performance over the last 30 days and explain why ROAS declined this week.” Manus then generates a written breakdown outlining key metrics, trend shifts, and potential contributing factors.

Natural language access to ad data

Graphed also highlights Manus’s ability to interpret campaign data using natural language. Traditionally, extracting insights requires working with filters, attribution settings, breakdown menus, and custom columns. More advanced analysis may require API exports or external reporting tools.

With Manus, advertisers can type questions directly. Prompts such as “Why did ROAS decline over the past seven days?” or “Which ad sets had the highest CPA last month?” replace multi-step reporting processes.

This conversational layer translates intent into data queries behind the scenes, reducing the need for technical knowledge of reporting structures.

Proactive anomaly detection

Another capability identified in testing is anomaly detection. Graphed reports that Manus can surface unusual shifts in campaign performance during analysis. These include fluctuations in return on ad spend, sudden cost changes, and delivery variations.

Rather than waiting for a direct question, Manus can flag irregular trends as part of its review of account data. This early pattern recognition may help advertisers identify performance changes sooner than manual reporting cycles.

Importantly, Manus does not correct issues automatically. It identifies signals that appear abnormal and presents them for review. Budget, creative, and targeting decisions remain with the advertiser.

Structured task execution

Manus is designed as an autonomous agent capable of breaking down multi-step tasks and executing them sequentially. Testing suggests it performs best when given clear, structured instructions.

When users provided detailed prompts with defined objectives, outputs were more organized and relevant. This indicates it is better suited to task-based workflows than open-ended brainstorming.

Where Manus struggles

Inconsistent access and external redirects

One major finding involves inconsistency in how Manus appears inside Ads Manager. In some accounts, clicking the tool redirects users to an external site rather than remaining fully embedded. According to Graphed, in those cases, users are taken to manus.im, asked to log in, and prompted to begin a subscription with a trial period. Some testers reported that the experience did not feel like a seamless native integration.

This suggests that the current rollout may not represent a uniform in-platform experience across all accounts.

Incorrect interpretations of Meta ad data

A central concern in Graphed’s analysis involves how Manus processes advertising data. The system was originally designed to work with structured data standards intended to reduce hallucinations.

However, Meta advertising data is delivered through complex, nested API structures. These include multiple layers of metrics, attribution details, and performance signals that are not always presented in simplified formats.

Graphed explains that when an AI system works directly with raw API outputs instead of receiving a validated semantic layer, interpretation errors can occur. Tester reports referenced in the article indicate that Manus sometimes produced incorrect interpretations during early experiments.

The issue is not whether Manus can access data, but whether that data is structured in a way that minimizes misinterpretation.

Access limits and eligibility requirements

Access to Manus inside Ads Manager remains restricted. According to Graphed, the tool currently works primarily with the Sales campaign objective and specific conversion setups.

Certain advertiser categories are excluded, including Housing, Employment, Credit, Social Issues, Elections, Politics, Financial Services, and Pharma. Some accounts are also experiencing delayed rollouts.

Additional limitations include spending thresholds, policy compliance requirements, desktop-only design, and Admin or Editor permissions. View-only users cannot access the feature. There is also limited public information regarding pricing, data retention policies, and detailed documentation about how the system interacts with advertiser dashboards.

Added automation without added transparency

The broader issue is transparency. Meta’s advertising ecosystem is often described as a “black box,” where algorithms determine bids, placements, and budget allocation with limited visibility.

The concern is that Manus operates on top of this same infrastructure. Rather than increasing transparency, it may introduce another automation layer. If decisions are made autonomously, efficiency may improve. However, advertisers could lose clarity over how outcomes are generated.

When automation increases faster than transparency, it becomes harder to diagnose performance shifts or understand the logic behind results.

Infrastructure limits constrain potential

Grpahed explains that while Manus may be technically capable of rapid task execution, Meta’s underlying API system imposes constraints on automated changes. For example, automated budget adjustments are capped per ad set within specific time windows. This limits how quickly optimization decisions can be applied.

As a result, Manus’s performance is shaped not only by its model design but also by platform architecture and system rules.

Built on an existing automation system

Manus does not function independently. It operates on top of Meta’s existing advertising automation framework, including tools like Advantage+ campaigns.

If underlying measurement or attribution systems contain inconsistencies, adding an AI interface does not automatically resolve them. Improved presentation does not correct foundational data limitations.

What this means for advertisers

The rollout signals continued movement toward embedded AI analysis inside campaign management tools. Key use cases focus on faster reporting, natural language querying, and anomaly detection.

However, structural questions remain regarding data handling, integration consistency, and infrastructure alignment.

The current phase appears to represent early deployment and testing rather than a fully standardized system-wide release. Performance and access may vary depending on account configuration and rollout status.

The development highlights both the potential and the constraints of integrating autonomous AI directly into advertising platforms.