Guide
AAAI vs IJCAI: one added an AI reviewer, the other added a $100 fee
The two have quietly diverged on the thing nobody compares: who or what reads your paper first. One is scaling review with machines, the other by making submission expensive and reviewing rewarding.
Published 14 August 2026
Still choosing between them? Paste the abstract lower on this page and see which rooms publish work like yours. Jump to the matcher
The short version
- The two are nearly identical on paper. Both want 7 pages plus references, both reject in a first phase where you have no right of reply, and their deadlines are half a year apart.
- What differs is who reads it first. AAAI gives Phase 1 three reviewers, two human plus one labelled AI review carrying no rating, and requires qualified authors to be available to review.
- IJCAI goes the other way. Two humans, reviewers forbidden from having an LLM assess anything, and 100 dollars a submission unless none of its authors is on another paper.
- The rejection rules are mirror images. AAAI rejects on reviews that are sufficiently negative, IJCAI on two that are not sufficiently positive. If your result is correct and unexciting, that is the whole comparison.
The published rules
Fifteen axes where the two venues actually differ
Every cell comes from a page one of the two conferences published. Where a venue says nothing, the cell says so.
| AAAI-27 | IJCAI-ECAI 2026 | |
|---|---|---|
| Reviewers in the first phase | Three: two human reviews plus one AI review carrying no rating and no recommendation. | Two PC members, both human. |
| What ends a paper there | Reviews that are sufficiently negative, with no opportunity for author feedback. | Two good-quality reviews that are not sufficiently positive, with no opportunity to submit an author response. |
| When that verdict lands | 24 September 2026, about eight weeks after the deadline, with all reviews released at once. | 4 March 2026, about six weeks after the deadline. |
| LLMs in the reviewing | One AI review on every main-track paper, plus an AI discussion summary for the senior committee only. Decisions stay human. | Reviewers may use an LLM only to polish their own prose: no generated opinions, assessments, judgements of quality or recommendations. |
| Second phase | More reviewers, up to five in total, who cannot read the Phase 1 reviews until they file their own. | PC and SPC members review, an Area Chair writes the meta-review, a Senior Area Chair cross-checks it. |
| Author response window | 19 to 25 October 2026, seven days. | 7 to 10 April 2026, four days. |
| What the response may contain | One rebuttal for all reviews, 2,500 characters, no URLs, no new results even if a reviewer asked. | Optional, one page on a supplied template. A clarifying figure is allowed; code links and new experiments are not. |
| Your duty as a reviewer | Qualified authors are required to be available to review, at a light load, unless extenuating circumstances apply. | By submitting, each author volunteers to be entered into the pool of potential PC members. |
| Submission fee | None on any page we read. | USD 100 per submission, waived only where no author appears on any other submission. Non-refundable even after a desk rejection. |
| Papers per author | 10 to the technical track. | 8. Extras are identified by submission ID and desk-rejected after the deadline. |
| Length | 7 pages of content, 9 in total, everything past page 7 references only. Two more were purchasable at AAAI-26 at $300 each. | 9 pages: 7 for the body, 2 for references. |
| Previous rejections | No disclosure field on the pages we read. | You must declare any rejection from a peer-reviewed conference in the last 12 months and supply that version with its reviewer comments. |
| Anonymity | Non-anonymised manuscripts are among the compliance failures that cut AAAI-26 from almost 29,000 submissions to roughly 23,000. | Anonymisation must cover the PDF metadata, and papers violating the rule are rejected without review. |
| Your own use of an LLM | AAAI-27 permits judicious use, authors responsible for plagiarism and invented references. AAAI-26 prohibited generated text, and no AAAI-27 author-policies page exists yet. | It does not satisfy the criteria for authorship and use must be documented. IJCAI says it will employ LLM detection tools and desk-reject papers found to be written by one. |
| If it is accepted | One author must register and present in person. | A six-minute talk with two minutes of questions, in person in Bremen, and an A0 poster. |
AAAI from its AAAI-27 review process page, main technical track call, submission instructions, AAAI-26 rebuttal FAQ and review process update. IJCAI from the IJCAI-ECAI 2026 main-track call, submissions FAQ, primary paper initiative and accepted papers page, plus the 2025 reviewers FAQ and rebuttal FAQ, the most recent reviewer-facing guidance it has published. Read August 2026. The columns are different editions because that is what exists.
One venue bought capacity, the other rationed demand
Two answers to one problem: submissions grew faster than reviewers. AAAI ran an AI review over all 22,977 of AAAI-26's full-review papers in under a day at under a dollar each, on donated API credits, and left every decision to humans. IJCAI charges 100 dollars on every submission unless it is the authors' only one, and pays the money back as registration waivers to good reviewers. They push authors in opposite directions.
The inversion
A lukewarm paper dies faster at IJCAI
The tinted row above repays a second reading. These are not one test at two heights. Only one of them is survivable by a paper nobody has an opinion about.
Not sufficiently positive is a much easier bar to fall below
IJCAI rejects a paper whose two Phase 1 reviews are not sufficiently positive, so two reviewers who find nothing wrong and nothing worth arguing about end it on 4 March, before you have written a word. AAAI rejects on reviews that are sufficiently negative, from an allocation of three. Indifference is fatal at one venue and survivable at the other.
So send the solid, unexciting result to the venue with more surface
This is where treating the two as interchangeable costs a paper. A correct incremental result nobody will call important has more room at AAAI: survive Phase 1 and it collects up to five reviews and a seven-day response window, and recommendations there are explicitly not made by averaging scores. A sharp, arguable claim is fine at either, because a reviewer with an opinion clears the IJCAI bar.
The calendar
Six months out of phase, which settles more submissions than prestige does
IJCAI has not published a 2027 calendar. Its last two cycles closed in the third week of January and notified in late April, so treat the shape as stable and the dates as the 2026 ones they are.
| Stage | AAAI-27 | IJCAI-ECAI 2026 |
|---|---|---|
| Abstract due | 21 July 2026 | 12 January 2026 |
| Full paper due | 28 July 2026 | 19 January 2026 |
| Payment for non-primary papers | No fee exists | 26 January 2026 |
| First-phase rejections | 24 September 2026 | 4 March 2026 |
| Author response | 19 to 25 October 2026 | 7 to 10 April 2026 |
| Final notification | 30 November 2026 | 29 April 2026 |
| Camera ready | 14 December 2026 | 20 May 2026 |
| Conference | 16 to 23 February 2027, Montreal | 15 to 21 August 2026, Bremen, jointly with ECAI |
From the AAAI-27 conference page and the IJCAI-ECAI 2026 important dates, both anywhere on earth, read August 2026. The IJCAI 2025 cycle ran a week later at every stage, which is why the shape is treated as stable. Take your dates from the current call.
AAAI answers seven weeks before IJCAI closes, and IJCAI wants the reviews
AAAI-27 notifies on 30 November 2026; IJCAI's last two paper deadlines were 19 and 23 January. It is the obvious route between the two and it leaves real time to revise. Nobody plans for the receipt: IJCAI requires you to declare any peer-reviewed rejection from the previous 12 months and attach that version with its reviewer comments. Write the resubmission as a document that arrives holding its own criticism.
What neither venue publishes
Nobody has an acceptance rate for these two, including the sites quoting one
The middle column is the entire published record. Everything else in circulation is somebody's arithmetic over a denominator the venue never gave.
| Figure | What the venue published | What circulates |
|---|---|---|
| IJCAI-ECAI 2026 acceptances | 990 papers across all tracks, 713 of them main track. | Rates computed from it, although the page gives no submission total. |
| IJCAI 2025 acceptance rate | None. IJCAI has never published one. | 17.6% and 19.28%, from trackers that disagree by nearly two points and cite no IJCAI source. |
| AAAI-26 submissions | Almost 29,000 to the main technical track, roughly 23,000 after compliance removals, 22,977 in the AI review pilot. | 4,167 of 23,680, quoted as 17.6%, on a denominator matching none of AAAI's own three counts. |
| How many papers the first phase kills | Not published, at either venue, for any edition. | AAAI's nearest figure: 98.98% of Phase 2 papers held three or more human reviews when responses opened. |
| The review scale | AAAI publishes no review form. IJCAI's last published one is from IJCAI-ECAI 2022 and used a named nine-point recommendation. | Numeric scales for both. Neither current form is public: AAAI has not posted one, and OpenReview returns a permission error to a guest. |
| An appeal after a first-phase rejection | No route on the IJCAI main-track call, submissions FAQ or principles page, and none on the AAAI-27 process page. | Nothing documented. That is not proof no back channel exists, but it is all anyone can honestly say. |
Counts from the IJCAI-ECAI 2026 accepted papers page, the AAAI-26 review process update and the AAAI-26 AI review pilot paper; the IJCAI-ECAI 2022 review form applies to no later edition. Read August 2026. The circulating figures are named so you can recognise them, not because we can verify them.
Free tool, no account
Where should this paper go?
Paste the title and the abstract. It ranks the venues that publish this kind of work, with the reach and the article charge OpenAlex has on record.
Both first phases end before you get to speak
At either venue the file you upload meets two reviewers with no right of reply, so nothing you would have said in a rebuttal helps. Which of the two rooms it goes to is the part still open.
Matching needs no account. A paid review does not predict a decision, and what it cannot do is written down. What ends a submission before review is on the pre-submission checklist.
phdflow reads a finished draft against the guidelines of the venue you are aiming at and returns an accept or reject call with the objections behind it. Scored against 697 real decisions: AUC 0.91 and 96% accuracy on 297 ICLR 2025 papers, one false accept in 149 rejects. The full measurement. Read a finished review before you pay for one. Your first paper gets a free preview; the full report is 60 credits, €9. No account needed. Pricing.