AudioEye Study Finds AI Coding Tools Do Not Write Accessible Code

AudioEye (AEYE) published research testing AI-generated websites for accessibility and a survey examining business confidence in AI-written code.

Rhea-AI Summary

AudioEye (AEYE) published research testing AI-generated websites for accessibility and a survey examining business confidence in AI-written code.

Five AI coding tools each built three websites using a brief requiring WCAG 2.2 AA, a web accessibility standard. Testing found 306 distinct accessibility issues, appearing more than 59,000 times across the 15 sites; 91% were rated medium or high severity. The generated sites averaged 55 issues per page, compared with 62 in AudioEye's 2026 Digital Accessibility Index. In the survey, 81% expressed confidence that AI-generated code met accessibility standards, while 73% reported increased accessibility issues and complaints since adopting AI.

News Explained

The survey adds a legal-claim measure: almost half of respondents had received a demand letter or lawsuit, and 71% of that group said AI tools were used on the pages receiving the complaint.

Key Figures

Websites tested: 15 websites Distinct accessibility issues: 306 issues Medium- or high-severity issues: 91% +5 more

Websites tested

15 websites

Five AI tools tested across three website types

Distinct accessibility issues

306 issues

Found across the tested websites

Medium- or high-severity issues

91%

Share of issues identified

Issues per page

55 vs. 62 issues

LLM-coded sites vs. the typical website in AudioEye's 2026 Digital Accessibility Index

Confidence in AI-generated code

81%

Survey respondents confident their site's code meets accessibility standards

Reported accessibility issues and complaints

73%

Survey respondents reporting an increase since adopting AI

Demand letter or lawsuit

46%

Survey respondents who had received one

AI tools used on complained-about pages

71%

Respondents who received a demand letter or lawsuit

Previous AI Reports

1 past event · Latest: Jun 25

Same Type 1 event

1. Jun 25

   Accessibility index

   24h Move

   -2.5%

   Index reported 62 accessibility issues per page, the benchmark cited in the current study.

24h Move is the share-price change in the day after each event; other market factors may also have contributed.

Key Terms

wcag 2.2 aa, llm

2 terms

wcag 2.2 aa technical

"sites that meet WCAG 2.2 AA requirements"

WCAG 2.2 AA is the mid-level conformance tier (Level AA) of the World Wide Web Consortium’s Web Content Accessibility Guidelines version 2.2. It is a technical standard made up of specific “success criteria” that web sites, web applications, and digital documents must meet to be considered accessible at AA; meeting it requires satisfying all Level A and Level AA criteria in the specification. The criteria cover things like perceivable alternatives, keyboard operability, readable contrast and focus visibility, predictable navigation, and other requirements intended to make digital content usable by people with a range of disabilities; it is a standards document, not a legal rule, though jurisdictions and contracts often reference it for compliance.

llm technical

"The LLM-coded sites averaged 55 issues per page"

A large language model (LLM) is an advanced computer system trained on vast amounts of written text to understand and generate human-like language, similar to a very fast, well-read assistant that can summarize documents, draft messages, or answer questions. Investors care because LLMs can speed up research, automate customer support, and reduce costs, while also creating new product opportunities and risks around accuracy, bias, and regulatory oversight that can affect a company’s performance.

AI-generated analysis. How Rhea-AI works. Not financial advice.

See more from StockTitan in Google Search and AI answers. Adds StockTitan as a preferred source · opens Google

Add on Google Not now

81% of developers trust AI-generated code for accessibility, but the data shows they should not

TUCSON, Ariz., Sept. 30, 2026 /PRNewswire/ -- AudioEye, Inc. (Nasdaq: AEYE) ("AudioEye" or the "Company"), an industry-leading digital accessibility company, today published new research and conducted a business survey that showed two critical findings: leading AI coding tools failed to build accessible websites, while 81% of teams using them believe that they are meeting compliance standards.

AI coding tools build inaccessible sites with high-severity issues

AudioEye gave five AI tools, including OpenAI, Anthropic, Google, xAI, and Lovable, the same brief: build a news and media site, an online store, and a financial services site that meet WCAG 2.2 AA requirements, the global standard for accessibility compliance.

Testing across the 15 sites found 306 distinct accessibility issues, appearing more than 59,000 times across the pages. A single mistake in a navigation menu or form field repeats on every page that uses it. Of the issues found, 91% were medium or high severity, meaning they make tasks harder or impossible for people with disabilities and AI agents. These are also the types of issues that show up consistently in digital accessibility lawsuits.

The LLM-coded sites averaged 55 issues per page, while the typical website in AudioEye's 2026 Digital Accessibility Index averaged 62. Asking an AI tool to build an accessible site made little difference; their pages were still nearly as inaccessible as the average website today.

AI did not hallucinate these failures. It has been trained on an inaccessible web, and it writes inaccessible code as fast as teams can ship it.

View full study →

The Trust Gap: Teams are confident in AI-generated code even as complaints climb

AudioEye also surveyed developers and business leaders on how much they trust AI-written code. The results show a gap between what they believe and what they are experiencing. While 81% say they are confident that the AI-generated code on their site meets accessibility standards, 73% report accessibility issues and complaints have increased since adopting AI.

Almost half (46%) have received a demand letter or lawsuit, and 71% of those respondents said AI tools were used on the pages receiving the complaint.

View full survey →

What this means for teams building with AI

None of this means AI and accessibility are incompatible. It means a tool trained on inaccessible web content cannot judge or fix its own accessibility. Organizations relying on LLMs alone for accessibility are accumulating legal exposure with every page they ship.

"AI is producing code faster than ever, and it's clear there is a false sense of security that AI tools write accessible, compliant code," said Kelly Georgevich, CEO of AudioEye. "AudioEye Intelligence, our proprietary dataset, is built on more than one million human reviews, billions of interactions, and years of testing with assistive technology. We have paired this intelligence with the latest in generative AI, allowing customers to ship accessible code at scale."

Methodology

AudioEye asked five AI tools (OpenAI, Anthropic, Google, xAI, and Lovable) to each build three websites that followed WCAG 2.2 AA standards. AudioEye then scanned and tested the resulting 15 websites for accessibility issues like keyboard focus, dialog handling, form error announcements, and screen reader output. Each issue was then rated high, medium, or low severity based on impact to the end user and legal risk.

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
Aulani Capuchin
aulani.capuchin@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-ai-coding-tools-do-not-write-accessible-code-302893917.html

SOURCE AudioEye, Inc.

FAQ

AI-generated questions and answers. How Rhea-AI works. Not financial advice.

How did AudioEye test AI-generated websites for accessibility?

AudioEye asked five AI tools to each build a news and media site, an online store and a financial services site meeting WCAG 2.2 AA standards. It scanned and tested the resulting websites for keyboard focus, dialog handling, form error announcements and screen reader output, then rated issues by severity.