AI Conference Deadlines and Disclosure Planning

The deadline is closer than the paper thinks

AI conference deadlines create a strange kind of pressure. The paper may feel nearly done. The disclosure often does not.

That gap matters. Many AI, ML, NLP, HCI, and data science venues now ask authors to say how they used AI tools. Some ask in the paper. Some ask in a submission form. Some ask during camera-ready upload. Some say little, then point authors to broader publication ethics rules.

If you wait until the final hour, you will write a weaker disclosure. You may forget which tool helped with code, which prompt changed the abstract, or whether an AI assistant touched reviewer rebuttal text.

This guide gives you a deadline plan. Use it alongside your normal conference tracker. Then generate an [AI Usage Card](/) before submission day, so your disclosure does not become a rushed paragraph at 23:47 AoE.

Use deadline trackers, but verify the venue page

Deadline trackers help you plan. They save time when you compare venues across AI, machine learning, NLP, computer vision, robotics, and related fields. One open source tracker describes itself as a countdown for AI, ML, NLP, and vision conference deadlines, sorted by what comes due first. (aideadlines.nauen-it.de)

Use those trackers as calendars, not as policy sources.

Before you make a submission decision, open the official call for papers. Dates move. Tracks differ. Workshops follow their own rules. A main track, dataset track, industry track, and workshop may all use different portals and deadlines.

For example, the ICML 2026 call listed a full paper submission deadline of January 28, 2026 AoE, while the NeurIPS 2026 call listed a full paper deadline of May 6, 2026 AoE. (icml.cc) ACL 2026 listed January 5, 2026 as the submission deadline through ARR and March 14, 2026 as the commitment deadline for ACL 2026. (2026.aclweb.org) ACML 2026 moved its conference track submission deadline from June 26 to July 5, 2026, a useful reminder that dates can change even after authors start planning. (acml-conf.org)

Write the official source into your project notes. Add the deadline time zone too. AoE sounds generous until your coauthor in another time zone uploads the wrong PDF.

Track two deadlines, not one

Most teams track the paper deadline. Fewer track the disclosure deadline.

Treat [AI disclosure](/ai-disclosure-in-systematic-reviews-and-meta-analyses/) as its own work item. It needs evidence, not memory. If you used AI for writing, coding, translation, figure drafting, data cleaning, screening abstracts, or reviewer response drafting, you need to know what happened and where it affected the work.

The easiest method is simple. Add one line to your project log each time AI changes the research workflow.

You do not need a diary. You need enough detail to answer a reviewer or editor without guessing.

A useful entry includes:

• the tool name and version, if available
• the date or phase of work
• the task
• the human check you performed
• whether outputs entered the manuscript, code, data, figures, analysis, or supplementary files

If that sounds like too much, generate an AI Usage Card early and update it as your project changes. The card gives you a record that you can paste into an acknowledgment, methods note, supplementary file, or submission form. For examples of wording, see [AI Usage Cards examples and templates](/ai-usage-cards-examples/).

Read the AI policy before you polish the paper

Conference AI policies do not all say the same thing.

ICML 2026 allowed authors to use generative AI tools such as large language models to assist with writing or research, according to its call for papers. (icml.cc) AAAI-26 took a stricter line for authored text: its policy prohibited papers that include LLM-generated text unless that text appears as part of the paper's experimental analysis, while still allowing LLMs for editing or polishing author-written text. (aaai.org) ICLR 2026 described a policy approach under which LLM use had to be disclosed and individuals remained responsible for their contributions. (blog.iclr.cc)

These differences change what you can do the night before submission.

One venue may accept grammar polishing with disclosure. Another may reject generated prose even if you edited it. A reviewer policy may also affect your rights as an author. ICML 2026, for instance, described two policies for reviewer LLM use and let authors choose whether their paper required a stricter review setting or allowed a more permissive one. (icml.cc)

So read the policy before the final writing sprint. If your conference has no AI policy, check its publisher, society, or submission system. If you still find no rule, use a transparent disclosure anyway. The question Do I need to disclose AI usage in my paper? gives a practical decision path.

A timeline that works for conference submissions

Use this schedule for any AI-related conference. It fits main tracks, workshops, shared tasks, and doctoral consortia.

Six to eight weeks before the deadline

Choose your target venue. Save the official call for papers, author instructions, formatting rules, anonymity rules, ethics rules, AI policy, and supplementary material policy.

Start the AI usage log.

If you already used AI before choosing the venue, reconstruct the record now. Go through your chat history, code commits, document comments, and lab notes. Do not wait until the paper reaches the final template.

Ask a blunt question: did AI touch the scientific contribution, or only the writing surface?

That difference will shape your disclosure. AI-assisted grammar editing usually needs a shorter note than AI-supported data extraction, code generation, image processing, or literature screening.

If you work with agents, read Do conference papers need to disclose AI agent use?. Agent workflows can blur the line between tool use and task delegation.

Four weeks before the deadline

Draft the disclosure before you draft the final abstract.

This feels backwards. It works.

When you write the disclosure early, you see missing checks. Maybe you used an AI tool to refactor analysis code but never reviewed the changed functions. Maybe a model suggested citations and you have not verified them. Maybe you used machine translation for interview excerpts and need to state who checked the translation.

Fix those gaps while you still have time.

At this stage, generate a draft card at ai-cards.org. You can revise it later. If your team uses Overleaf, the guide on using AI Usage Cards in Overleaf shows how to keep the card near the manuscript. If you write in LaTeX, the LaTeX tutorial for AI Usage Cards gives more options.

Two weeks before the deadline

Map your AI record to the venue format.

Some conferences want an acknowledgment. Others want a statement in the limitations section, ethics section, reproducibility checklist, or submission form. A few want no identifying acknowledgment during anonymous review, but allow one in the camera-ready version.

Do not break anonymity to disclose AI use. Instead, write a neutral statement that does not identify authors, institutions, grants, labs, or unpublished systems.

A blind-review-safe version might look like this:

\section*{AI assistance disclosure}
The authors used AI-based writing assistance to improve grammar and clarity in selected parts of the manuscript. The authors reviewed and edited the affected text and take responsibility for the final content. No AI tool was used to generate research claims, citations, data, or experimental results.

If your venue forbids unnumbered sections or asks for disclosure elsewhere, adapt the wording. The content matters more than the heading.

For journal-like conference submissions, see AI transparency requirements for journal submissions. Many conference proceedings now borrow journal-style ethics and authorship language.

One week before the deadline

Freeze the AI workflow.

This does not mean you must stop using every tool. It means you stop adding new AI-supported tasks without recording them.

Late changes create disclosure problems. A coauthor may ask an assistant to shorten the conclusion. Another may ask a coding tool to modify a plot script. Someone may use AI to draft a rebuttal plan before reviews arrive.

Write a team rule for the final week:

% Team note for final submission week
% Record any AI assistance before merging text, code, figures, or supplementary files.
% Include tool, task, date, and human verification.
% Do not paste AI-generated citations or claims without source checks.

This small rule prevents confusion. It also protects the corresponding author, who often signs the submission form.

Submission day

Prepare three versions of the disclosure.

First, keep the full AI Usage Card for your records. Second, prepare a short manuscript statement. Third, prepare a plain text version for the submission portal.

A short statement might read:

\section*{AI assistance disclosure}
The authors used an AI writing assistant to edit author-written text for grammar and concision. The authors also used an AI coding assistant to suggest unit tests for analysis scripts. All suggested text and code were reviewed by the authors. The authors verified the final manuscript, code, citations, and results.

For portal fields, remove LaTeX commands:

The authors used an AI writing assistant to edit author-written text for grammar and concision. The authors also used an AI coding assistant to suggest unit tests for analysis scripts. All suggested text and code were reviewed by the authors. The authors verified the final manuscript, code, citations, and results.

If the venue asks whether AI generated any part of the submission, answer according to its definitions. Some policies separate editing, translation, code, figures, and generated prose. Do not compress everything into "writing help" if the tool affected analysis.

Do not disclose like this

Weak disclosure usually has one of two problems. It says too little, or it shifts responsibility away from the authors.

Avoid vague lines like:

The authors used AI tools during the preparation of this manuscript.

That sentence tells the reader almost nothing. It does not name the task. It does not say whether AI touched claims, code, data, or language. It does not say who checked the output.

Also avoid wording that makes the tool sound responsible:

ChatGPT improved the manuscript and ensured accuracy.

A tool cannot ensure accuracy. Authors can check sources, rerun analyses, inspect code, and accept responsibility.

Better:

The authors used ChatGPT to suggest edits for grammar and sentence structure in the introduction and discussion. The authors reviewed the edits, checked all claims and citations, and take responsibility for the final manuscript.

If you need more examples, use How to disclose ChatGPT usage in academic papers and How to choose and disclose AI tools for research.

Special cases: rebuttals, reviews, and camera-ready files

Conference work does not end at submission.

You may use AI during rebuttal, camera-ready revision, poster preparation, video scripts, or artifact cleanup. Check the policy again before each phase. Some venues publish separate rules for authors and reviewers. Some track policies differ.

If you review papers, treat AI use with extra care. You may handle confidential manuscripts, unpublished results, and author identities. The guide on AI disclosure in peer review covers what reviewers and editors should report.

If your paper gets accepted, update the AI Usage Card before camera-ready upload. The version you submit at camera-ready should reflect the final manuscript, not the first submission. If AI helped write the rebuttal but none of that text entered the paper, record it in your internal card and disclose it if the venue asks about the review process.

A simple deadline checklist

Use this checklist when you add an AI conference deadline to your calendar.

First, save the official call for papers. Then save the AI, ethics, anonymity, authorship, supplementary material, and camera-ready rules.

Next, start an AI usage log. Do this before the writing sprint.

Then generate a draft AI Usage Card. Share it with coauthors. Ask them to add missing tool uses.

Two weeks before the deadline, convert the card into the format the venue accepts. One week before the deadline, freeze new unrecorded AI use. On submission day, paste the short version into the manuscript or portal, and keep the full card with your project files.

That is the whole system. It is not glamorous. It works because it moves disclosure out of memory and into the research record.

Generate your AI Usage Card at ai-cards.org before the next conference deadline. Use it as your submission note, your LaTeX disclosure, or your internal record for the camera-ready version.