AudioEye Study Finds up to 68% Drop in AI Agent Task Completion on Inaccessible Websites
An independent check using WebJudge agreed with AudioEye’s scoring 95% of the time.
Rhea-AI Summary
AudioEye (AEYE) published research on September 24, 2026, finding lower AI-agent task completion on websites without accessibility fixes.
The study tested 1,560 AI agents built from six commercial models across 13 tasks on six websites. On the site with the most accessibility issues, agents completed 31% of tasks without fixes versus 96% on its accessible version. Across all tests, median token use was 128,000 without fixes versus 90,000 with fixes, a 43% increase. On five websites, the sites themselves were unchanged; the test varied whether AudioEye’s fixes loaded. The sixth was a demonstration site with accessible and inaccessible versions.
Positive
- None.
Negative
- None.
Details
News Market Reaction – AEYE
On Sep 24, the day this news came out, AEYE closed 0.84% above the previous close. Our momentum scanner recorded 3 alerts in the available session data.
Data tracked by StockTitan Argus for the Sep 24 session.
Key Figures
- AI agents tested
- 1,560 agents
- Study methodology across six websites
- Maximum task-completion drop
- Up to 68%
- Inaccessible websites
- Task completion
- 31% vs. 96%
- Inaccessible site vs. accessible version of the same site
- Average token use
- 43% more tokens
- On inaccessible sites
- Median run token use
- 90,000 vs. 128,000 tokens
- Accessible vs. inaccessible conditions; the article reported 43% more compute without fixes
- Failed attempts
- 60 of 60 attempts
- Six models tested a task involving numbers shown only in an image without a text description
Key Terms
accessibility tree technical
AI-generated analysis. How Rhea-AI works. Not financial advice.
Agents also used up to 6x more tokens on inaccessible sites. Digital accessibility makes a site compliant and AI-ready.
On the site with the most accessibility issues, agents completed just
AI agents read a webpage using its accessibility tree, a digital map of the webpage's content. When the map is incomplete, the agent cannot finish what it came to do. Tasks fail, and the customer never gets the answer, the booking, or the purchase they sent the agent to complete.
Key findings
- Task completion fell by up to
68% on inaccessible websites. Agents completed96% of tasks on an accessible site and31% on an inaccessible version of the same site. Accessibility was the only difference between the two versions. - Agents used
43% more tokens on average on inaccessible sites, and up to 6x at the extreme. Across all 1,560 tests, the median run consumed43% more compute when accessibility fixes were absent (90,000 vs. 128,000 tokens), a gap AudioEye calls the compute tax. On the site with the most accessibility issues, every model used at least 2x as many tokens without accessibility fixes, and the highest-cost model used 6x as many. - Missing accessibility labels stopped every model tested. For example, one task needed numbers that appeared only inside an image, and that image had no text description in the code. No agent could read them. Six models tried it ten times each, and all 60 attempts failed. That is the problem in one example: if the code is inaccessible, the software has nothing to read.
AI readiness starts with accessibility
In September 2026, Meta launched Muse, a personal AI agent designed to act on users' behalf, including sending emails, filling out forms, booking travel, and making purchases. Meta's family of apps reaches more than 3.6 billion people daily, giving the company significant reach as AI agents become more common, alongside products such as Instinct and Grokbot.
"The number of agents using the internet is expected to increase exponentially. However, agents use a website's accessibility tree, not the DOM, which powers web browsers. A core part of AI readiness in the future will be having an accessibility solution, not just for compliance, but for the agents themselves," said Kelly Georgevich, CEO of AudioEye. "Our research shows that even the most advanced AI model can fail outright on an inaccessible page. The accessibility investment our customers make for compliance turns out to be the groundwork for the agentic web, and AudioEye is built to make websites AI-ready at scale for businesses of all sizes."
Companies build accessible websites to meet compliance standards and to serve people with disabilities, and both reasons hold as strongly as ever, driven by increasing litigation. The research adds a compelling third reason: an accessible website is legible to any software that arrives, whether a screen reader or an AI agent sent by a customer. Businesses that invest in accessibility now are building accessible websites for the customers they have today and the AI agents they will send tomorrow.
The full report is available at www.audioeye.com/post/ai-readiness-starts-with-an-accessible-website/.
Methodology
AudioEye ran 1,560 independent AI agents built from six commercial models across 13 real-world tasks on six websites. Every agent started fresh, and each test was repeated ten times. On five websites in retail, travel, and direct-to-consumer commerce, the only change was whether AudioEye's accessibility fixes loaded. The websites themselves were never modified. The sixth site is the W3C's demonstration site, published in both accessible and inaccessible versions.
Results were independently rechecked using WebJudge, an open-source evaluation system from Ohio State University's NLP group, which agreed with AudioEye's scoring
About AudioEye
AudioEye exists to ensure the digital future we build is accessible for people with disabilities and AI agents. AudioEye's comprehensive solution for over 129,000 customers, combines industry-leading AI automation technology with expert fixes informed by the disability community. This powerful combination delivers industry-leading protection against lawsuits and readiness for AI agents. With 25 US patents, AudioEye's solution includes 24/7 accessibility monitoring, automated WCAG issue testing and fixes, expert testing, developer tools, and legal protection, empowering organizations to confidently create accessible digital experiences for all.
Media Contact
Mike Barton
press@audioeye.com
Investor Contact
Tom Colton
Gateway Group, Inc.
AEYE@gateway-grp.com
949-574-3860
View original content to download multimedia:https://www.prnewswire.com/news-releases/audioeye-study-finds-up-to-68-drop-in-ai-agent-task-completion-on-inaccessible-websites-302888585.html
SOURCE AudioEye, Inc.
FAQ
AI-generated questions and answers. How Rhea-AI works. Not financial advice.
How did website accessibility affect AI-agent task completion in AudioEye’s study?
On the site with the most accessibility issues, agents completed 31% of tasks without fixes, compared with 96% on an accessible version of the same site.
What happened when AudioEye tested AI agents on an image without a text description?
All 60 attempts failed. The task required numbers shown only in the image, which had no text description in the code; six models each tried the task ten times.
How were AudioEye’s AI-agent study results checked?
The results were independently rechecked with WebJudge, an open-source evaluation system from Ohio State University’s NLP group. It agreed with AudioEye’s scoring 95% of the time.