Navigating arXiv's New AI Policy: How to Avoid a One-Year Ban
Introduction
Scientific publishing is facing a new challenge: the influx of AI-generated papers filled with fake citations, nonsensical diagrams, and unedited prompt responses. The preprint server arXiv, a cornerstone for physics, mathematics, computer science, and related fields, has taken a strong stance. In a recent announcement, Thomas Dietterich—emeritus professor at Oregon State University and a key member of arXiv’s editorial advisory council and moderation team—revealed that any submission containing inappropriate AI-generated content will result in a one-year ban from submitting new papers. Even after the ban lifts, all future submissions must undergo peer review before arXiv will host them. This guide walks you through the policy, what it means for researchers, and how to ensure your work complies.
What You Need
- Understanding of arXiv’s policies: Familiarize yourself with the latest guidelines on AI-generated content.
- Awareness of common AI slop: Recognize fake citations, unedited prompt responses, and nonsensical diagrams.
- Original research or review work: Ensure your manuscript is authored by humans and properly vetted.
- A clean track record: Previous violations may escalate consequences.
- Peer review access: If you’ve been banned, you’ll need a plan for getting papers reviewed before posting them on arXiv.
Step-by-Step Guide to Compliance
Step 1: Understand What Qualifies as AI-Generated Slop
Before you submit, know what the system is looking for. arXiv defines inappropriate AI-generated content as material produced entirely by large language models or other generative AI without meaningful human oversight. Common red flags include:
- Fake citations: References to nonexistent papers or authors.
- Unedited prompt responses: Text that reads like a chatbot output, often with repetition or lack of logical flow.
- Nonsensical diagrams: Figures that don’t accurately represent data or contain gibberish labels.
- Hallucinated facts: Claims or equations that aren’t grounded in reality.
If your paper contains any of these, it will be flagged as slop and you’ll face the ban.
Step 2: Review the Official Announcement
Stay updated by checking arXiv’s official channels. Thomas Dietterich’s social media thread is a key source, but arXiv may publish formal guidelines. Look for announcements on their blog or moderation page. The policy applies across all fields hosted on arXiv, including physics, mathematics, computer science, quantitative biology, and statistics.
Step 3: Audit Your Submission for AI Use
If you’ve used AI tools (like ChatGPT, Claude, or GitHub Copilot) during writing or research, review the final manuscript carefully. Ensure:
- Every citation is real and checkable.
- All figures are original and meaningful.
- The text flows naturally and doesn’t sound like a raw AI response.
- No section was generated by AI without your full editorial control.
arXiv’s moderation team will scan submissions for patterns typical of AI slop. Even if you used AI as a helper (e.g., for grammar or summarization), you must take responsibility for the final content.
Step 4: Submit with Transparency (If Applicable)
While arXiv doesn’t currently require a disclosure statement for AI use, being transparent can help. Some journals now ask authors to declare AI assistance. For arXiv, the safest approach is to avoid any reliance on generative AI for core intellectual contributions. If you do mention AI use, be prepared to explain how you vetted the output.
Step 5: Prepare for Peer Review If You’ve Been Banned
If you’ve already submitted AI slop and received a ban, here’s what the one-year period entails:
- No new submissions: You cannot post any papers to arXiv.
- Permanent requirement: After the ban, every future manuscript must undergo peer review (e.g., from a journal or conference) before arXiv will post it. This means you need to secure a peer-reviewed acceptance first—even for preprints.
Use the ban year to improve your research ethics. Consider submitting your work to traditional journals with rigorous peer review to rebuild your reputation.
Step 6: Monitor for Updates
Policies evolve. Check arXiv’s help page or follow their social media for changes. As of now, the one-year ban and permanent peer-review condition apply to all fields. If you’re unsure, reach out to the moderation team before submitting.
Tips for Avoiding Trouble
- Never submit raw AI output. Even if you think it’s well-written, it often contains subtle errors that humans miss.
- Use AI as a tool, not a co-author. Always critically check every sentence and citation.
- Keep your research notes, code, and data. If arXiv’s moderators question your submission, you can provide evidence of human work.
- Stay informed about community standards. Other preprint servers and journals are likely to follow arXiv’s lead.
- If you’re a student, consult your advisor. They can help ensure your paper meets ethical guidelines.
By following these steps, you can protect your ability to use arXiv while maintaining scientific integrity. Remember: a one-year ban is not just a punishment—it’s a signal that the research community takes AI-generated slop seriously.