Academic peer review has always been a conversation among experts, but the question on Reddit about ICLR reviewers using AI tools reveals that the conversation itself is changing faster than many are ready to admit. Our view is straightforward: leaning on AI to assist with review work is not a shortcut, it is an inevitability that the community must learn to manage with transparency and intention. The tension here is not about technology versus tradition; it is about how openly we acknowledge the tools we already use.
This discussion connects directly to what we have seen in our own coverage. When we explored How to write papers that connect with top conference audiences, the underlying challenge was clear: researchers are adapting their writing to be clearer and more compelling in a crowded field. Now that same adaptive pressure has reached the review side. If authors are expected to produce polished, accessible work, reviewers are naturally exploring every available method to keep up with their own workload. Meanwhile, our piece on The Real Cost of Choosing Speed Over Skill in AI Research warned that rushing can undermine genuine expertise. That warning applies equally here: using AI to summarise a paper is one thing, but relying on it to form a critical judgement without deeper engagement risks the very foundation of peer review.
The practical question for every reviewer reading this is not whether you should use AI, but how you should use it and whether you will be honest about doing so. A researcher who asks an AI to check for methodological flaws or to surface related work is working smarter, not cutting corners. But a reviewer who feeds a paper into a model and copies the output verbatim is abdicating responsibility. The conference organisers at ICLR, like leaders at top venues everywhere, now face a governance problem. They need standards that acknowledge real-world behaviour, because pretending reviewers are not using these tools only pushes the practice underground, where it cannot be discussed or refined.
Watch for how ICLR handles this in the coming submission cycle. If they address AI-assisted review head-on with clear disclosure expectations, they will set a template for the entire field. If they remain silent, the gap between public norms and private practice will widen, and the quality of feedback will suffer accordingly. The goal is not to ban the technology, but to ensure that every review still represents a human being who has read, thought, and decided. That standard is worth defending, and the next few months will show whether the community is ready to do so.