The submission volume for AAAI-26 has skyrocketed: Nearly 30,000 papers, with 20,000 from China, and the review system is on the verge of collapse.

You may not believe it, but the AAAI-2026 conference you submitted to should have the largest number of submissions in history.

Previously, the number of applicants exceeded 30,000, and many of them switched from NeurIPS.

Now the official data is also public: The main technical track received nearly 29,000 submissions, and nearly 20,000 submissions were from China, accounting for an astonishing two-thirds.

Source: https://aaai.org/conference/aaai/aaai-26/review-process-update/

The number of paper authors has also skyrocketed. This time, more than 75,000 independent authors submitted papers. What does this mean? It means that even if only 1% of the authors ask questions at a certain moment, the reviewers will need to reply to about 750 emails, which is enough to overwhelm a conference run by volunteers.

Another astonishing figure is that in the past few months, the team has received more than five times the total number of emails in the whole year of AAAI-25. At the peak, there were up to 400 email requests per day.

However, AAAI doesn't accept all papers. They will also reject some substandard papers, such as those without a PDF, non-anonymous manuscripts, papers exceeding the page limit, and papers with more authors than the submission limit. Even so, approximately 23,000 papers still entered the review process, almost twice the number of reviewed papers in AAAI-25 (12,957)!

In addition to the number of submissions, AAAI-26 also announced the top three research keywords, which are computer vision (nearly 10,000 papers), machine learning (nearly 8,000 papers), and natural language processing (more than 4,000 papers). It seems quite difficult to stand out among these submissions, as everyone's research fields are concentrated in these areas.

As the number of submissions continues to rise, the challenges faced by the review system are also intensifying, especially when the scale of papers reaches tens of thousands. This has pushed the review system to its limits in terms of storage, computing, bandwidth, process support, and the most scarce resource - the time of qualified reviewers.

To meet this demand, AAAI recruited more than 28,000 program committee members, senior program committee members, and area chairs. The scale of the program committee of AAAI-26 is almost three times that of AAAI-25.

You don't have to worry. Although the number of submissions has increased sharply, the quality and fairness of the review will still be guaranteed. AAAI is also taking some measures, specifically as follows:

AAAI is actively investigating possible ethical issues in the review process. For confirmed violations, corresponding consequences will be pursued, and these consequences may extend beyond the current review cycle and even impose sanctions in the more distant future. In addition to the ethics chair of AAAI-26, AAAI also has a publication committee and an ethics committee to continue investigations and impose sanctions outside the scope of this conference.

In addition, the AI-assisted review experiment has shown positive early results, including tools for detecting and resisting collusion among reviewers.

The matching of papers and reviewers uses state-of-the-art algorithms and sets up robustness checks to prevent bid manipulation. This means that unauthorized mutual bidding has little impact on the paper matching process. In the paper matching process of AAAI-26, bidding is just one of many considerations, and more important factors include research expertise areas, previous publications, and geographical diversity.

Given the extremely large number of submissions and the imbalance between the number of reviewers and papers in each sub-field, some reviewers will review papers adjacent to their core research fields rather than papers that perfectly match their expertise.

Even so, there will still be a short delay in the review process. After all, the number is there. AAAI hopes that the paper authors can understand.

The comment section is also having a heated discussion about this staggering number of submissions.

Some people said it was expected, as most participants in the US Olympiad are from China.

There is also a very interesting comment.

It can't help but recall a widely circulated saying before: Now the global AI competition is a PK between Chinese-Americans and Chinese people.

Of course, some comments pointed out in a half-joking tone that behind this is the huge publication pressure well-known in the current AI academic circle. Actively submitting papers may also be a helpless move to "find a way out" for one's academic future in this "involution" environment.

There have been many discussions about the problems and controversies reflected by the sharp increase in the number of submissions to top AI conferences. For details, see the report by Almost Human:

Do some papers really do this? Papers from many top global universities actually hide AI praise instructions.

Who’s Adam? The most incredible NeurIPS review is out.

Is there a problem with the top AI conference model? The vicious cycle of "publish or perish" is crushing the entire AI academic circle.

Rejection warning: The way of "stealthily churning out papers" with large models is blocked. The strictest new rules of ICLR are here.

China's AI Behind the Data

This trend sweeping top conferences is by no means accidental. Multiple reports show that China and Chinese scholars are playing an increasingly important role in the field of artificial intelligence.

A report by Digital Science in July this year showed that China has become the absolute leader in the global artificial intelligence research field. It has not only surpassed other countries in the number of research but also in research quality (citation attention) and influence. Moreover, China has become the strongest partner of the United States, the United Kingdom, and the European Union in the field of AI research.

Address: https://www.digital-science.com/blog/2025/07/new-report-shows-china-dominates-in-ai-research/

An article on Medium more intuitively showed the amazing rise of this "Chinese power" in the past decade. Data showed that the proportion of papers with Chinese scholars' participation has skyrocketed in top AI conferences such as CVPR, NeurIPS, and ICML.

Address: https://medium.com/data-science-collective/a-decade-of-change-chinas-rise-in-ai-research-and-the-global-talent-flow-d9c49ebd4d37

Taking the top computer vision conference CVPR as an example, the proportion of papers by Chinese authors increased from about 30% in 2015 to nearly 40% in 2019 - 2020, surpassing the United States. In NeurIPS and ICML, this proportion also increased from a negligible 10% in 2015 to 20 - 30% in 2024.

The estimated percentage of accepted papers co - written by Chinese researchers. These numbers have increased significantly in recent years.

In 2024, among the top 20 research institutions in terms of the number of accepted papers at top global AI conferences (such as NeurIPS), 8 were from China, including the top - ranked institution.

Behind this achievement is the joint promotion of top universities such as Tsinghua University and Peking University, as well as technology giants such as Huawei, Alibaba, and Tencent.

This article is from the WeChat official account “Almost Human” (ID: almosthuman2014). The author is someone who pays attention to AI. It is published by 36Kr with authorization.