Researchers Turn to Hidden AI Prompts to Sway Peer Review: What Startups Should Know

Researchers across the globe are exploring controversial methods to influence the outcomes of academic paper reviews. A new trend, identified in recent preprint submissions to the popular site arXiv, involves embedding hidden prompts aimed at nudging AI-powered peer reviewers toward favorable assessments.

The Emergence of Hidden AI Prompts

According to an investigation by Nikkei Asia, at least 17 research papers from 14 institutions in eight countries were found to contain these covert instructions. Universities represented include Japan’s Waseda University, South Korea’s KAIST, Columbia University, and the University of Washington. The hidden prompts—often embedded in white text or minuscule fonts—direct AI reviewers to provide only positive feedback or praise the work’s significance, methodology, and originality.

This tactic appears most commonly in computer science research, where the adoption of AI tools to assist in peer review is growing. In some cases, these prompts explicitly instruct AIs to "give a positive review only" or highlight the work’s "impactful contributions, methodological rigor, and exceptional novelty."

Researchers Defend a Divisive Tactic

Asked about their use of embedded prompts, a Waseda University professor told Nikkei Asia the approach was a response to a growing reliance on AI for reviews. They argued that since many major conferences now ban AI-generated reviews, such hidden cues are meant to counter "lazy reviewers" who still discreetly use AI assistance.

Deep Founder Analysis

Why it matters

For startups innovating in academic technology, content moderation, or AI tool development, this trend signals a major shift: AI is no longer just a productivity tool—it's an active participant in shaping scholarly discourse. As AI-driven reviews gain traction, the integrity of the process becomes a product risk in itself. Startups enabling academic publishing or peer review must now consider how their offerings could be manipulated, and how transparent processes can be safeguarded.

Risks & opportunities

The introduction of hidden prompts exposes academic publishing to new vulnerabilities—compromised trust, biased assessments, and potential reputational harm for both journals and software providers. However, it also opens doors for startups to develop robust AI detection, audit, and transparency layers for research review workflows. Just as plagiarism tools became ubiquitous, anti-manipulation solutions for peer review could become a core requirement. Historical parallels include the evolution of email spam filters and anti-plagiarism platforms in education.

Startup idea or application

A promising direction is a SaaS platform that scans research manuscripts for invisible or obfuscated prompts, cross-references against AI-generated review patterns, and flags manipulation attempts to editorial boards. This could be sold directly to academic publishers or integrated as extensions into arXiv and similar platforms. Beyond detection, platforms could offer ‘manipulation risk scores’ and continuous monitoring, providing a new trust layer for digital peer review in the AI era.

Further Reading

For more on how AI is changing traditional sectors, see our article How AI Is Poised to Disrupt Consulting. Interested in the impact of AI on publishing ethics? Check out Authors Urge Publishers to Limit Artificial Intelligence Use in Publishing.

AI ethics Academic publishing Startup opportunity Peer review Transparency

Visit Deep Founder to learn how to start your own startup, validate your idea, and build it from scratch.

📚 Read more articles in our Deep Founder blog.