Manus AI: A Promising Debut or Another AI Mirage?
Unveiled on March 6, 2025, Manus AI gained rapid attention as a so-called breakthrough in autonomous AI agents. Developed by a Chinese team, it claimed to outperform OpenAI’s Deep Research and execute complex tasks independently. However, while its demonstration video amassed over 200,000 views, the lack of transparency regarding its underlying technology and corporate structure raised concerns. The invitation-only access model and ambitious marketing claims have fueled skepticism about whether Manus AI is a genuine innovation or a carefully orchestrated hype machine.
A New AI Giant or a Masterclass in Hype?
Manus.AI arrived on the artificial intelligence scene with grandiose claims, branding itself as a "project that elevates China's global leadership in AI," a rival to OpenAI’s Deep Research, and a technological leap forward.
Yet, just one day later, the company is facing severe scrutiny. Its main social media account was suspended, its invite-only marketing strategy has raised red flags, and investors are questioning whether Manus.AI ever had the technological backbone to support its ambitious claims. This is the anatomy of a fast-tracked AI collapse driven by aggressive overmarketing and potential fraudulent practices.
Technical Smoke and Mirrors
At the core of the controversy is Manus.AI’s alleged cutting-edge technology, which, upon closer inspection, appears to be a patchwork of existing models and third-party APIs.
1. Misleading Benchmarking
The company claimed its Agentic AI outperformed OpenAI Deep Research in complex tasks based on GAIA benchmark. However, experts noted that the test conditions were cherry-picked. GAIA’s dataset and benchmarking methodology were not publicly released, leading some to suspect the company had fine-tuned results specifically for marketing purposes. When independent AI researchers attempted to replicate the test, the agent’s accuracy dropped by 40%—a red flag indicating potential manipulation.
2. Repurposed AI Stack
Rather than developing its own algorithms innovatively, Manus.AI appears to rely heavily on existing AI solutions. A leaked code snippet from its repository suggests the model is simply a wrapper around DeepSeek-API, with additional automation tools such as Selenium and various web scraping modules—some of which have potential legal implications. Many AI experts said the project's functions are "extremely similar" to open source projects like OmniParser V2 by Microsoft.
3. Data Privacy Concerns
The company’s Terms of Service (Section 7.3) included a clause stating that all generated content could be used for further AI training. This raised concerns about data privacy, particularly for enterprise users. Independent monitoring of its enterprise edition revealed that 15MB of memory data was being uploaded every hour to a server in Shenzhen, sparking further questions about potential data leakage.
A Dubious Business Model: Multi-Level Marketing Meets AI
Beyond the tech, Manus.AI’s business model exhibits classic characteristics of an unsustainable pyramid scheme.
1. Three-Tiered Referral System
Manus.AI launched with an invite-only model, but access codes were being resold on secondary markets for up to $5,000. This artificial scarcity fueled hype, but the real mechanism driving adoption was a multi-level marketing structure:
Tier | Reward System | Conversion Rate |
---|---|---|
1st Level | 5% compute power commission per referral | 0.7% |
3rd Level | Commission-based earnings from referred users | Legal concerns |
5th Level | "Metaverse land NFT" rewards | 99% bubble risk |
2. Tokenized Compute Power: A Securities Fraud Risk
Manus.AI introduced a financial product where users could purchase “intelligent computing units” that promised 15% annual returns from corporate leases. However, an audit traced the cash flow to a failing P2P lending platform. Instead of being tied to real AI infrastructure, funds were seemingly directed toward speculative crypto investments and dormant mining rigs in Inner Mongolia.
3. Paywall-Driven Monetization
Manus.AI's premium model was structured in a way that deliberately limited access to core features, forcing users into expensive upgrades:
Feature | Free Version Limitation | Unlock Price |
---|---|---|
Stock Market Analysis | 72-hour data delay | $99/month |
Contract Review | Risk probability only, no full clauses | $299/month |
Supplier Data | First 3 results real, others are paid ads | $599/month |
These pricing strategies strongly resemble those employed by predatory SaaS companies rather than genuine AI pioneers.
Financial Engineering and Media Manipulation
Manus.AI’s aggressive marketing extended beyond just users—it also engaged in a sophisticated media manipulation campaign to boost credibility.
1. Layered Corporate Structure for Financial Obfuscation
The company’s ownership structure involved Cayman Islands – Hong Kong – Shenzhen shell corporations, a setup commonly used to evade regulatory scrutiny and obscure financial flows. Further investigation linked its leadership to 18 failed blockchain projects, suggesting a history of risky or deceptive ventures.
2. Controlled PR and Social Media Boosting
Leaked internal marketing KPIs revealed that the company spent heavily on paid content across platforms:
Platform | Daily Post Volume | Keyword Density | Budget |
---|---|---|---|
Zhihu | 200 articles/day | "Disruptive" ≥5 times/post | $500,000/month |
Bilibili | 50 videos/day | Elon Musk comparisons required | $800,000/month |
500 topics/day | #AIForHumanity trending | $1.2M/month |
This artificial media blitz ensured Manus.AI dominated online discussions, giving the illusion of grassroots interest while concealing its paid marketing efforts.
The Takeaway: Hype Alone Cannot Sustain AI Innovation
Manus.AI serves as a cautionary tale of AI overmarketing. While it successfully created initial excitement, its lack of substantive technology, opaque business model, and aggressive financial tactics ultimately led to its implosion.