AccessAgent 2027
Accessibility for the Agentic Web
Building an inclusive future for AI agents that navigate, reason, communicate, and act on the Web—across diverse abilities, interfaces, and contexts.
About the workshop
The Web is undergoing a fundamental transition from a collection of pages and applications navigated directly by users to an increasingly agent-mediated ecosystem. Recent advances in large language models (LLMs), multimodal foundation models, and autonomous AI agents have enabled systems that can browse websites, retrieve information, complete forms, execute transactions, interact with online services, and make decisions on behalf of users. Emerging agentic platforms are beginning to transform how people search, shop, learn, access healthcare, consume content, and interact with digital services. Rather than manually navigating complex interfaces, users may increasingly delegate web tasks to intelligent assistants capable of understanding goals, planning actions, and interacting with multiple online resources autonomously.
This transformation creates unprecedented opportunities for accessibility. AI agents can potentially act as personalized accessibility intermediaries, adapting web content to individual needs and abilities. For a blind user, an agent can summarize visual content, navigate complex websites, and present information in screen-reader-friendly formats. For Deaf or hard-of-hearing users, agents can provide multimodal communication support, generate real-time captions, or facilitate sign-language-aware interactions. For DeafBlind users, agents can coordinate communication across haptic, tactile, and braille interfaces while simplifying complex web workflows. Similarly, individuals with cognitive disabilities, learning difficulties, motor impairments, age-related challenges, or temporary situational limitations may benefit from agents that reduce cognitive load, automate repetitive actions, explain complex information, and provide adaptive assistance tailored to user capabilities.
However, the rapid emergence of agentic systems also introduces a new generation of accessibility challenges. Traditional web accessibility research has largely focused on webpage design, content structure, assistive technology compatibility, and compliance with standards such as WCAG. In contrast, AI agents operate as decision-making intermediaries between users and the Web. Accessibility is therefore no longer solely a property of the interface being displayed. It becomes a property of the entire interaction pipeline, including user intent understanding, task planning, tool selection, information retrieval, multimodal communication, action execution, explanation generation, consent management, and error recovery. An accessibility barrier at any stage in this pipeline can prevent users from fully benefiting from agent-mediated web interactions.
Furthermore, many current AI agents are designed and evaluated primarily around effectiveness, efficiency, and task completion. Accessibility considerations often remain secondary concerns evaluated after system deployment. Existing benchmarks rarely measure how agents perform for users with diverse sensory, cognitive, communication, or physical abilities. Key questions remain unanswered: How should agents communicate uncertainty to users with different accessibility needs? How can users understand and override agent decisions? How should agents personalize interactions without reinforcing stereotypes or making incorrect assumptions about user capabilities? What forms of explanation are most effective for users relying on assistive technologies? How can accessibility requirements be incorporated into agent planning and reasoning mechanisms rather than merely into front-end interfaces?
The challenge becomes even more significant as agentic systems increasingly interact with high-stakes domains such as healthcare, education, employment, banking, transportation, and government services. In these settings, inaccessible agent behavior may result not only in usability issues but also in exclusion, reduced autonomy, privacy violations, diminished trust, and unequal access to digital opportunities. If accessibility considerations are not incorporated early into the design, evaluation, and governance of Web agents, the community risks recreating existing barriers or introducing entirely new forms of digital exclusion at the very moment when agentic technologies are reshaping the future of the Web.
Despite growing research activity in web accessibility, assistive technology, foundation models, responsible AI, human-computer interaction, and autonomous agents, these communities largely operate independently. Accessibility researchers frequently focus on interfaces and assistive technologies, while AI-agent researchers primarily investigate planning, reasoning, tool use, and autonomy. Similarly, Web researchers, disability advocates, standards organizations, industry practitioners, and policymakers often engage in separate discussions despite addressing overlapping challenges. Consequently, there is currently no dedicated venue that systematically examines accessibility throughout the complete lifecycle of AI agents operating on the Web.
Important Dates
Mark your calendar for the workshop paper submission and camera-ready milestones.
All dates are stated as provided by the workshop organizers. Submission deadlines are end-of-day Anywhere on Earth (AoE). The workshop-day schedule will be announced separately.
Call for Papers
We invite original research, position papers, emerging ideas, empirical studies, and practical perspectives on accessibility throughout the lifecycle of AI agents operating on the Web.
Topics of interest include, but are not limited to:
Accessible Web agents
Accessibility-aware planning, task execution, assistive technology compatibility, accessible automation, and error recovery.
Human–agent interaction
Accessible autonomy, user control, consent, transparency, explanation, and communication of uncertainty.
Multimodal accessibility
Interactions spanning text, speech, images, video, sign languages, braille, haptics, and alternative interfaces.
Evaluation & benchmarking
Ability-aware agent benchmarks, interaction traces, usability barriers, equitable performance, and accessible task success.
Personalization & inclusion
Capability-aware adaptation, cognitive and motor accessibility, participatory design, and inclusive user modeling.
Trustworthy agentic Web
Fairness, privacy, safety, governance, responsible AI, and alignment with Web accessibility standards.
Paper format and submission
AccessAgent 2027 invites two types of contributions: short papers and long papers. Authors should carefully select the most appropriate category. Papers already published or simultaneously under review elsewhere are not eligible for archival publication in the workshop proceedings.
Original and highly promising research or applications. Submissions are assessed for originality, importance and technical quality, rather than breadth or maturity alone.
Substantive original research techniques, findings or applications. Authors should clearly position the contribution, explain innovative aspects, present sufficiently detailed reproducible methods, communicate findings and discuss broader implications for agentic Web accessibility.
The two categories differ primarily in the scope of their contributions, not their importance or technical quality. Both categories will undergo the same review process and may be selected for oral or poster presentation. Rejected long papers will not normally be reconsidered as short papers, except in rare cases where reviewers unanimously recommend this.
Formatting and supplementary materials
Please consult the official WWW 2027 workshop guidelines and the relevant submission track for applicable ACM formatting requirements. Optional supplementary materials may accompany a submission in PDF or ZIP format; their deadline is the same as the main paper deadline (4 January 2027).
Publication
Eligible accepted workshop papers may be included in the Companion Proceedings of The Web Conference 2027, subject to the conference's publication policies and camera-ready requirements. Publication is not through IEEE Xplore.
Submission link · OpenReview
OpenReview will be used for submission and peer review. All listed authors must maintain up-to-date OpenReview profiles, including current and previous institutional affiliations and conflict-of-interest information. Select the appropriate AccessAgent workshop submission track when it becomes available; the link below currently points to the overall ACM Web 2027 OpenReview venue.
Visit OpenReview submission portal →Organizers
An interdisciplinary team bringing together research in accessibility, Web technologies, multimodal AI, and human-centered intelligence.
Dr. Tirthankar Dasgupta
Senior Scientist with research spanning natural language processing, multimodal AI, Web intelligence, sign language technologies, and inclusive intelligent systems. His work includes assistive technologies for people with visual, hearing, and motor impairments.
Dr. Manjira Sinha
Lead Scientist in Accessibility Research, specializing in Web technologies, Web accessibility, assistive technology, natural language processing, and inclusive AI. Her work focuses on improving Web access for people with diverse disabilities.
Prof. Mufti Mahmud
Distinguished Professor and researcher in cognitive computing and brain informatics, with interests in explainable, multimodal, privacy-preserving, and inclusive AI for healthcare, ageing, and education.
Professor David J. Brown
Director of the Computing and Informatics Research Centre and Research Group Leader of the Interactive Systems Research Group, Department of Computer Science, School of Science and Technology, Nottingham Trent University.