As of February 2021, JAIR publishes, on a regular basis, detailed metrics on the submission handling process. This transparent publishing approach is intended to provide useful information to prospective authors and valuable calibration to the editorial team and to reviewers.
Submission evaluation statistics are based on data from 2022/06/16 to 2023/06/16, released on 2023/09/14. These statistics do not cover the 90 days previous to the date of release in order to minimize distortion from submissions currently being evaluated.

A submission received on 16 June 2022 would have been accepted by 31 October 2022 with a probability of 24.9%, on average, if not summarily rejected by 3 July 2022.

In case of a rejection with encouragement to resubmit, assuming resubmission happens 6 weeks after the first decision (i.e., by 12 December 2022), the revised version would have been accepted by 7 April 2023, with a probability of 63.8%, on average.

This implies an overall acceptance probability of 19.4% (52.6% if not summarily rejected) and an expected time to acceptance (possibly after one revision) of 42.1 weeks, i.e., less than 10 calendar months.

The times mentioned above are averages. Especially the times required for the peer-reviewing process can vary considerably, depending on the length and quality of the submission in question, and on the load and availability on the Associate Editors and reviewers with the relevant expertise. Longer submissions often take longer to evaluate, as do submissions that fall into narrowly defined subject areas in which the potential pool of qualified reviewers is very small. Above-average processing times are also more likely to occur during the reviewing periods of very large conferences in AI, such as AAAI, IJCAI, ICML and NeurIPS, when many potential reviewers tend to have high work loads.
A significant fraction of first-time submissions are summarily rejected by the Editor-in-Chief, the Associate Editor-in-Chief, or an Associate Editor, when they do not fall within the scope of the journal, as stated in the Editorial Charter. In particular, submissions are often rejected without being peer-reviewed if they fail to present research results of importance to the global artificial intelligence community, if they fail to satisfy the technical submission requirements, or if they describe specific applications of AI, but fail to provide methodological advances or insights of interest to the AI research community; the last criterion does not always apply to special tracks on particular application areas of AI. 
JAIR is a top-tier journal in AI. Therefore, papers that have been rejected from a conference tend to be summarily rejected from JAIR, unless their authors have achieved major improvements in quality and scope. Chapters of theses are also typically unsuitable for publication in JAIR without major rewriting and, in the context of surveys derived from thesis chapters covering related work, without providing major added value, by means of novel and insightful organisation of a research area within AI. Authors who have not previously published in JAIR are urged to read a sample of JAIR articles to get a feeling for the kinds of contributions that are expected by the journal and its readership.
JAIR does not accept submissions subject to major revisions. Instead, submissions that are deemed to likely be suitable for publication in JAIR after major revisions are rejected with encouragement to resubmit. Revised versions of such papers are handled whenever possible, and nearly all cases, by the same Associate Editor and reviewers, in order to ensure consistency of the evaluation process and to prevent authors to have to deal with a moving target with respect to the expectations of reviewers and editors. Resubmissions tend to be handled faster and tend to be accepted with much higher probability than first-time submissions.
Once accepted, articles tend to be published rapidly, unless authors are slow to respond to editorial and technical requests during the production process.