Accesserty Observations
Why it is hard to answer “does this US website have an accessibility statement?”
We looked for accessibility statements on 3,437 websites used in everyday life in the United States. This article explains the US rules for statements, how we searched, and why “does it have a statement?” depends on how you search, which sites you can reach, and what you count as a statement.
Citation details
- Published
- Updated
- Source snapshot date
Accesserty (2026-11-06). Why it is hard to answer “does this US website have an accessibility statement?”. Source snapshot date 2026-11-04. https://accesserty.com/en/observations/us-accessibility-statements-hard-to-find/
Summary
Whether a statement is required depends on the jurisdiction and the organization. Federal agencies must maintain an accessibility statement under a memorandum from the White House Office of Management and Budget (OMB), and some states have their own rules; Colorado, for example, requires state and local governments to post one. The Department of Justice’s 2022 guidance on business websites does not require a statement. We also found no public source that collects US accessibility statements in one place, so we had to search site by site.
A page called “Accessibility” is not always about the website. The first run did not check whether the content was about digital accessibility and found 1,395 records; the second run required it to be about website or app accessibility, and 818 remained. Even pages about the website are not always statements: of the 764 pages we could analyze, 488 were dedicated website accessibility statements according to AI page-type labels.
Not found does not mean absent. Our rule-following crawler could not get into, or could not read, 1,765 sites. Their absence from the data does not mean they have no statement.
Key figures
- Candidate websites
- 3,437
- Related pages collected
- 796
- 839 records, deduplicated by statement URL
- Dedicated accessibility statements
- 488
- of 764 analyzable pages, by AI page-type labels
- Unreachable or unreadable
- 1,765
- 1,442 unreachable or refused, 323 robots.txt; does not mean no statement
Taiwan has badges. What does the US have?
In Taiwan, one way to see whether a website works on accessibility is to check whether it holds an accessibility badge from the Ministry of Digital Affairs. Badge records are public, and our observation on Taiwan private accessibility badge records is based on that data.
In the US, we found no comparable nationwide public lookup. A website that wants to explain its accessibility usually publishes an accessibility statement, describing the standard it follows and how to get in touch about problems.
So we asked what looked like a simple question: do the websites people use in everyday life in the US publish accessibility statements? Once we started, it turned out to be anything but simple.
US rules: who has to publish a statement?
US rules on website accessibility fall roughly into three groups, depending on who runs the site. Below we cover federal rules and the state examples we found; we did not survey every state.
Federal agencies:
- Under the Section 508 standards, federal agency websites must meet WCAG 2.0 Level A and AA.
- OMB Memorandum M-24-08, issued in December 2023, requires federal agencies to maintain an accessibility statement on their websites (providing the information directly or linking to it), covering the standard used, known limitations where applicable, contacts, feedback channels, how to file a complaint, and the date it was last updated or reviewed, among other items.
State and local governments:
- In 2024 the US Department of Justice issued a rule under Title II of the Americans with Disabilities Act (ADA) requiring state and local government websites and apps to meet WCAG 2.1 AA.
- In April 2026 the Department extended each deadline by about one year in an interim final rule: from April 24, 2026 to April 26, 2027 for governments with a population of 50,000 or more (other than special districts), and from April 26, 2027 to April 26, 2028 for smaller governments and all special district governments.
- The rule does not require an accessibility statement. In its explanation the Department notes that governments can provide one to tell the public how to report problems, citing the W3C guidance on statements, but it declined suggestions to mandate specific practices such as statements or third-party audits.
- Some states have their own rules. Colorado’s technology accessibility rules (8 CCR 1501-11), for example, require state and local governments to post a technology accessibility statement with minimum content, and California requires the director and chief information officer of each state agency or entity to post on the home page, every two years, a signed certification that the website meets accessibility standards.
Businesses:
- The Department’s 2022 guidance says the requirements of ADA Title III for public accommodations, such as stores, banks, and restaurants, also apply to the goods and services they offer online. But the Department has no regulation setting detailed technical standards for these websites; WCAG and Section 508 are offered as helpful guidance. The guidance does not require businesses to publish an accessibility statement either.
For comparison, under Directive (EU) 2016/2102 the European Commission set a model accessibility statement for public sector websites in 2018, requiring the statement to give the conformance status, the content that does not conform, a feedback mechanism, and more.
In other words, among the rules we found, federal agencies and government bodies in some states must publish a statement, with differing content requirements; for other websites, we found no rule requiring one. Statements follow no common format, and that is where the first difficulty comes from.
What we wanted to answer: websites used in everyday life
We did not try to answer what share of US websites have a statement; we cannot, for reasons explained below. We wanted to look at the websites a person runs into while living everyday life in the US.
We organized candidate websites into categories by the tasks people need to get done, and after collection mapped the categories to three layers:
- Essentials: government services, health insurance, banking and credit cards, telecom and utilities, transport and travel, shopping, groceries and pharmacies.
- Frequent or high-impact: education and childcare, housing, jobs, cars and parcels, weather and local news, taxes and personal finance, and more.
- Social participation and quality of life: tickets and entertainment, restaurants and food delivery, long-term care, community organizations, fitness, pets, and more.
In total we put together 3,437 candidate websites. The largest categories were tickets and entertainment (640), education and public services (576), and government services (485).
| Layer | Main categories (candidate websites) | Candidate websites | Analyzable pages | Dedicated statements |
|---|---|---|---|---|
| Essentials | Government services (485), health insurance (313), transport and travel (248), banking and credit cards (170), shopping (132), telecom and utilities (106), groceries and pharmacies (83) | 1,569 | 335 | 259 |
| Frequent or high-impact | Education and public services (576), jobs (93), cars and parking (84), housing (67), parcels and moving (67), local news (50), weather (38) | 1,038 | 304 | 169 |
| Social participation and quality of life | Tickets and entertainment (640), restaurants and food delivery (131), software tools (14), fitness (12) | 830 | 125 | 60 |
| Total | 36 categories | 3,437 | 764 | 488 |
Each website’s layer was assigned after collection, based on its category. The layers follow the scheme Accesserty uses for other countries’ data; categories that scheme lacks, such as restaurants and food delivery, were placed by the closest existing category. When several sites share one page, the first record’s category is used. “Analyzable pages” and “dedicated statements” are explained below.
To be clear, we compiled this candidate list ourselves; it is not a random sample:
- 3,340 were compiled by hand, 45 came directly from a public ranking of popular websites, and 52 were added from earlier collection records.
- 1,474 candidates came with a statement URL from our earlier data, which we fetched again as hints; 697 were found this way, while other hints were blocked by robots.txt or failed to load. In other words, some sites are on the list precisely because we already knew they had a statement.
So the data speaks for “the websites we observed,” not for all US websites.
Difficulty 1: no place to look it up, so we searched site by site
We found no public source that collects statements, so we wrote a program that visited each of the 3,437 websites. It checked the hint URL first, then looked through homepage links, common URLs, and sitemaps. The run took place between November 3 and 4, 2026 (Taipei time). Of the 818 records it found, 697 were confirmed through hint URLs.
The program followed each site’s robots.txt, the file sites use to tell automated programs where not to go, as defined in the Internet Engineering Task Force (IETF) standard RFC 9309. Following Accesserty’s existing inclusion rules, sites that load an overlay tool, or host their statement on an overlay vendor’s domain, were not included.
Correction (2026-11-07): The collection program checked only the page source, so it missed overlays that scripts add after a page loads. We later loaded the 488 statements analyzed in the next article in a browser and checked them again. 449 were checked; the other 39 could not be, because they failed to load, robots.txt disallowed them, or the site refused the request. Among the pages checked, 7 are on sites that load an overlay: 6 on sites with a commercial overlay, and 1 on a site where a website-management service adds an assistive toolbar. Under the inclusion rules, these 7 should not have been included. The figures here are the results of the original collection and were not recalculated. The 7 are 1.4% of the 488; their effect on the next article’s shares is described in that article’s method notes.
| Result | Websites |
|---|---|
| Website or app accessibility page found | 818 |
| Could not connect or was refused | 1,442 |
| Connected, but nothing found | 671 |
| Disallowed by robots.txt, or robots.txt unreadable | 323 |
| Page exists but excluded by inclusion rules (overlay) | 181 |
| Already in other certification data, skipped | 2 |
| Total | 3,437 |
Difficulty 2: a page called “Accessibility” is not always about the website
On the first run, the program did not check whether the content was about digital accessibility: any accessibility page found through hint URLs, homepage links, common URLs, and so on was collected, for 1,395 records. When we checked them, many were not about the website at all but about physical access: how to book wheelchair seating, where accessible parking is, whether assistive listening devices are available on site. That information matters to visitors, but it is not a website accessibility statement.
So on the second run we added a condition: the page’s main content (excluding menus, headers, and footers) had to mention at least one of the following to count as being about website or app accessibility:
- WCAG
- Section 508 or EN 301 549
- Digital, website, or app accessibility
- Screen readers
- Assistive technology
The two runs used the same candidates and took place on November 3–4 (Taipei time). The result went from 1,395 records to 818. The difference has two sources: the new condition, and changes in fetch results between the runs, such as a site that could not be reached the second time.
| Category | Run 1: no digital-content condition | Run 2: with the digital-content condition |
|---|---|---|
| Tickets and entertainment | 330 | 84 |
| Travel and transit | 102 | 34 |
| Education and public services | 308 | 249 |
| Government services | 210 | 168 |
| All categories | 1,395 | 818 |
What was excluded? 579 sites appeared in the first run but not the second: 297 had no qualifying page in the second run, 277 could not be reached, and 5 had other outcomes. We randomly sampled 60 of the 579, fetched the pages from the first run again (still following robots.txt), and checked the content with keywords:
- Of the 40 pages we could read, 26 used only words about physical facilities, such as wheelchairs, parking, elevators, and service animals, and none about website accessibility.
- 11 used neither kind of words.
- 3 used words about website accessibility; 2 of them also used words about physical facilities.
- The other 20 had too little main text to judge.
So most of the readable excluded pages used only words about physical facilities. But a few pages that mention website accessibility did not make it into the second run, and sites whose content could not be read dropped out too. The 818 figure may therefore be low, which is why we later added a manual check.
With the same sites, a different condition and a different run took the result from 1,395 to 818. When a site has an “Accessibility” page, check what it is actually about.
Difficulty 3: not found does not mean absent
Many of the 3,437 sites were out of our program’s reach:
- The server refused the request (HTTP 403): 726 sites.
- robots.txt could not be read: 285 sites, including connection failures, no response within 15 seconds, or a file over 500 KiB. Under RFC 9309, when robots.txt is unreachable because of network or server errors, a crawler must treat the whole site as disallowed. Stopping on timeouts and oversized files was our own cautious choice; RFC 9309 only requires parsing at least 500 KiB.
- robots.txt returned a server error or another error code: 11 sites, also treated as fully disallowed.
- Stopped because robots.txt explicitly disallowed us: only 27 sites.
In other words, of the 323 sites where robots.txt finally stopped the program, only 27 explicitly said “stay out”; far more often the program could not even read robots.txt.
We could have done it differently, for example by pretending to be an ordinary browser or ignoring unreadable robots.txt files. But we kept the program within the rules. Respecting a site’s settings for automated programs matters more than collecting a few more records.
The cost is that some sites are missing from the data. Their absence does not mean they have no statement.
Adding a manual check: 839 records collected
The program has limits. Some sites generate their content with JavaScript, so the program sees a blank page; some block the program but open normally in a browser. We re-checked, one by one, the statements we had previously found by hand:
- 21 records were added after manual checking, on top of the 818 the program found; 20 of them were confirmed in a browser.
- That makes 839 records in total.
- None of these 21 are among the 488 statements analyzed in the next article: when fetched again, 17 returned HTTP errors and the other 4 did not meet the main-text length or English threshold.
About 30 more sites are not included for now, for example because they would not open, robots.txt disallowed us, the page no longer exists, or the site uses an overlay. We will check them one by one later.
A second reading: 488 dedicated statements among 764 pages
The collection rule was that a page relates to website or app accessibility. But a page about website accessibility is not necessarily a statement. To prepare the content analysis in the next article, we read every page again:
| Step | Count |
|---|---|
| Collected records | 839 |
| Deduplicated by statement URL: a page shared by several sites (for example, agencies of the same state government) counts once | 796 |
| Analyzable when fetched again on November 4, 2026 (readable, at least 300 characters of main text, in English) | 764 |
| Dedicated website accessibility statements by AI page-type labels | 488 |
| Page type | Pages |
|---|---|
| Physical or service accessibility, with the website as one section | 89 |
| Institutional accessibility hub linking to training, policies, forms, and more | 74 |
| Nondiscrimination or ADA notice | 25 |
| Disability services for students or patients | 23 |
| Guidelines for staff or content authors | 15 |
| Formal accessibility policy or regulation | 11 |
| Product accessibility features | 10 |
| Unfilled template or empty page | 9 |
| News, blog posts, and other | 20 |
Schools stand out: of 238 education pages, only 109 were dedicated statements. Many schools’ “Accessibility” pages are resource hubs or student services pages.
Page types were labeled page by page by AI against a fixed definition. We randomly sampled 60 pages and had a different AI model judge them again without seeing the original labels: the two models agreed on 54. We reviewed the other 6 by hand, and the original label was right on 3. The 54 agreed pages were not checked by hand, so 488 is the number of pages the AI labeled as dedicated statements, not a human-verified figure.
Conclusion
We set out to answer a simple question, and most of the effort went into deciding what counts as an answer.
In the US, whether a statement is required depends on the jurisdiction and the organization, statements follow no common format, and we found no central place to look them up. So the answer to “does this website have an accessibility statement?” depends on how you search, which sites you can reach, and what you count as a statement.
This also shows the value of public records from another angle: centralized, public data such as Taiwan’s badge records makes “does it have one?” a question you can look up. In the next article, we open the 488 statements and look at what they actually say.
Method
- Accesserty collected the data itself between 2026-11-03 and 2026-11-04 (Taipei time). The 3,437 candidate websites were organized by everyday-life category. The crawler followed robots.txt (RFC 9309), and pages the crawler could not reach were checked manually.
- The condition for “related to website or app accessibility” is described above. The difference between the two runs was checked with keywords on a random sample of 60 pages.
- The three layers were assigned by category after collection; they were not recorded at the time of collection.
- Page types were judged page by page by an AI model (Claude) against a fixed definition, reading each page’s title, main heading, and main text; for long pages, only about the first 2,800 and last 700 characters. A random sample of 60 pages was then judged again by a different AI model (OpenAI Codex) without seeing the original labels. The two agreed on whether the page is a dedicated statement for 54 pages; Accesserty reviewed the other 6 by hand, and the original label was right on 3. The 54 pages where the two models agreed were not checked by hand, and both could be wrong on the same page, so this check cannot be used to estimate the accuracy of the original labels.
- This article publishes aggregate results only. It does not provide record-level data, name or rank individual websites, or quote individual statements.
Limits
- The candidate list was compiled by us, not randomly sampled, and some sites are on it because we already knew they had a statement. The data cannot tell what share of US websites have an accessibility statement.
- When we found nothing, it may only be because we could not get in or read the page. The data cannot show that a given site has no statement.
- Statements are written by the sites themselves. We did not test any website, and having a statement does not mean a site is accessible.
- Page types were labeled by AI. In the 60-page check, the two models disagreed on 6 pages, and the original label was wrong on 3 after human review. Pages where the models agreed were not checked by hand and could still both be wrong.
- The rules section covers federal rules and the state examples we found; we did not survey every state.
Sources and related pages
- Next: US accessibility statements, promise or proof?
- Taiwan private accessibility badge records
- OMB Memorandum M-24-08 (2023-12-21)
- GSA Section508.gov: Website accessibility statement (updated 2025-03)
- U.S. Department of Justice Title II web rule, Federal Register 2024-07758 (2024-04-24)
- U.S. Department of Justice Title II compliance date extension, Federal Register 2026-07663 (effective 2026-04-20)
- ADA.gov: Fact sheet on the Title II web rule
- U.S. Department of Justice: Guidance on Web Accessibility and the ADA (2022-03-18)
- Colorado Office of Information Technology: guide to the state technology accessibility rules (8 CCR 1501-11)
- California Government Code § 11546.7 (state agency website accessibility certification)
- U.S. Access Board: Revised 508 Standards (published 2017-01-18)
- Commission Implementing Decision (EU) 2018/1523: model accessibility statement (2018-10-11)
- IETF RFC 9309: Robots Exclusion Protocol (2022-09)
- All observations