A browser window with a crossed-out image puzzle and a form, with a protective shield with a checkmark in front and the blue part of the SiteCockpit logo behind it on a light blue circle.

Accessible CAPTCHA: Alternatives to Image Puzzles in Shops

CAPTCHAs are meant to stop bots but often lock people out. Making a CAPTCHA accessible works best without a test for users: this article covers the rules, the alternatives and the checks for your shop.

November 19, 2026 • WCAG & Technology • 16 mins read

Summary

Under WCAG 1.1.1, a CAPTCHA is only acceptable if it offers another form for other types of perception. For sign-in, WCAG 2.2 (3.3.8, level AA) bans cognitive tests without an alternative, and image selection is exempt at AA. In shops, section 19 BFSGV also requires security functions to be operable. Fewest barriers arise with spam protection that needs no task from users, such as honeypots, filters and rate limits. As of 11/19/2026.

A CAPTCHA is meant to keep bots out of your accessible forms, but it also gives customers a test that some of them cannot pass. In an online shop that has consequences: anyone who cannot finish registration, login or a contact request does not buy and cannot report problems. Accessibility here means designing protection against spam and bots so that it locks nobody out.

This article explains what WCAG 2.2 and Germany's Accessibility Strengthening Act (BFSG) require for the accessibility of CAPTCHAs. It shows the common types and their barriers, presents options that need no test for users, and gives five checks for your shop.

Why CAPTCHAs exclude people

CAPTCHAs have been among the biggest obstacles on the web for years. In the WebAIM survey of screen reader users (no. 10, December 2023 to January 2024, 1,539 valid responses), the CAPTCHA ranked first among problematic items by a clear margin. According to WebAIM, this ranking has stayed largely the same for 14 years, and respondents with disabilities were twice as likely to rate the CAPTCHA as problematic as respondents without.

The W3C working group on accessible platforms describes the cause in its draft on CAPTCHA accessibility (editor's draft of April 13, 2022). For blind users, the very first step fails, because a screen reader cannot read out an image of distorted characters. People with a visual impairment or dyslexia also struggle to decipher them. Deaf and hard-of-hearing users are asked to type what they hear, and a rule of several attempts can overwhelm people with anxiety disorders. Time is a cost too: the draft cites Cloudflare data of 32 seconds on average per CAPTCHA.

The affected groups are larger than they seem. In the 2026 WebAIM survey (July to August, 1,780 valid responses, not a controlled sample), 83.2 percent reported blindness and 7.5 percent deafness or hearing loss. 115 respondents (6.5 percent) were both blind and deaf or hard of hearing. For this group, neither the image variant nor the audio variant gets them through. For blind customers, the order ends before it begins. The problem therefore does not affect a few users on the margins but the accessibility of the whole sign-in path.

A scenario: an online shop protects registration with an image selection ("Select all images with traffic lights"). From the perspective of a blind customer, the screen reader just reads "image", and an audio clip to type does not help. She gives up and buys elsewhere, and the shop owner never notices. Bots, on the other hand, can often solve such puzzles, according to the W3C draft. The problem stays invisible, and the question of the damage stays open.

There is also a security argument: according to the W3C draft, classic CAPTCHAs based on images, logic puzzles or audio can now be cracked largely by algorithms. Protection that no longer reliably stops bots but still holds up real customers is a poor trade.

Accessibility of CAPTCHAs: what WCAG and the BFSG require

WCAG allows CAPTCHAs if there is a second way through, and for sign-in, WCAG 2.2 no longer allows a cognitive function test without an alternative. Two criteria work together here.

Criterion 1.1.1: another form of the test

Criterion 1.1.1 (level A) requires a text description that names the purpose of the element, plus other forms of CAPTCHA for other types of perception. A CAPTCHA is an automated test meant to tell a human from a computer. A text alternative containing the solution would be pointless, since any computer could then pass. According to the W3C explanation of WCAG 2.1 (updated September 6, 2026), two different forms ensure that most people with disabilities find a form they can use.

Even so, every test remains unsolvable for some people. The W3C therefore recommends three further steps: offer more than two forms, allow a human customer service route that bypasses the test, and do not ask signed-in users for a CAPTCHA at all.

Criterion 3.3.8: accessible sign-in

Criterion 3.3.8 (level AA) is new in WCAG 2.2. No step of the sign-in may require a cognitive function test, such as remembering a password, typing characters or solving a puzzle, unless one of four exceptions applies. The W3C lists them: another sign-in method without such a test, a mechanism that helps (for example a password manager or pasting), recognizing objects, and recognizing your own content. It applies to signing in to an existing account, not to creating an account. A challenge that only appears after failed attempts or with an ad blocker active still counts.

Three consequences matter for operators. First, an audio CAPTCHA you have to transcribe does not meet the "alternative" exception. Second, arithmetic and logic puzzles are not exempt. Third, image selection ("Select all images with cars") is exempt at level AA but not at level AAA (3.3.9), and the W3C advises avoiding it where possible. A two-factor code also fails if users must type it and cannot paste it. You can find the new criteria at a glance on our page Understanding and implementing WCAG 2.2.

How the BFSG classifies it

For shops, section 19 no. 2 of the BFSG Ordinance (BFSGV) expressly names identification, authentication, security and payment functions. They must be perceivable, operable, understandable and robust. Under section 2 no. 3 BFSGV, a security function is any function that protects the security of information technology. A CAPTCHA protects a form against automated access and, in our assessment, can be understood as a security function. We have not found a statement from an authority or court specifically on CAPTCHAs (as of 11/19/2026).

What remains open is which version of EN 301 549 is the benchmark. Version V4.1.1 of September 2026 builds on WCAG 2.2 and therefore on 3.3.8. According to AccessibleEU, the European Commission's centre (notice of September 7, 2026), it is not yet the legal reference standard until the Commission cites it in the Official Journal. Until then, V3.2.1 based on WCAG 2.1 remains the reference. We could not confirm a citation as of 11/19/2026. The rule from 1.1.1 therefore already applies with WCAG 2.1, while 3.3.8 comes with WCAG 2.2: if you are planning now, meet both. You can read more about the standard in our article EN 301 549 V4.1.1: what changes for online shops.

Which types of CAPTCHA are problematic

Every test with a task excludes certain groups. The table classifies the common types based on the W3C sources.

Types of CAPTCHA, affected groups and classification under WCAG
TypeWho failsClassification
Typing distorted charactersBlind and visually impaired people, people with dyslexiaCognitive function test, not exempt under 3.3.8, plus another form required under 1.1.1
Audio to transcribeDeaf, hard-of-hearing and deafblind people, noisy surroundingsDoes not count as an alternative under 3.3.8
Arithmetic and logic puzzlesPeople with dyscalculia or cognitive impairmentsNot exempt under 3.3.8
Image selectionBlind and visually impaired peopleExempt at level AA, not at AAA (3.3.9), 1.1.1 still requires a second way
Sliders and gamesAnyone working without a mouse or touchscreenUsable only with several input methods
Protection in the backgroundNobody through a testNo cognitive function test, check privacy and older devices

The table also shows that exempt at level AA does not mean accessible to everyone. If you only meet the minimum, you still exclude people. Do not rely on behavior you tested once either: according to the W3C draft, a provider may fall back to classic CAPTCHAs depending on cookies, browser and assistive technology. The question is therefore not whether you use CAPTCHAs but what they cost your customers, and that also applies elsewhere, such as at checkout.

Alternatives: spam protection without a test for users

For accessibility, protection against spam and bots works best when it runs in the background. These are your options:

  • Honeypot: A hidden field that bots fill in and visitors never see; the W3C draft mentions it too. A human leaves the field empty, and the server discards every spam submission where it is filled. It must be unreachable for keyboard and screen reader, and the password manager must not fill it, otherwise you create a new barrier. According to OpenReplay, this solution is not enough on its own: advanced bots recognize the field, and targeted attacks by humans bypass it.
  • Spam filters on the server: Filters recognize typical patterns in the content. According to the W3C draft, with good tuning they can replace classic CAPTCHAs without burdening users.
  • Rate limits and timing checks: Limited submissions per device and session, plus rejecting forms that were filled out suspiciously fast. According to the developer blog, most bots submit a form instantly or within one to two seconds, and a check for under about three seconds catches many of them. According to the W3C, no method is 100 percent effective, but they can reduce the number of tests shown.
  • Proof of work: The browser solves a calculation in the background. Users barely notice it, it makes every submission more expensive for the operators of bots, and it may slow older devices.
  • Assessment without interaction: Some providers check in the background whether a computer or a human is acting. The W3C draft welcomes this for accessibility but points out that it often evaluates a lot of user data. Clarify this question with your data protection advisor.
  • A human route: A contact option through which a human bypasses the test is expressly recommended by the W3C. It helps above all blind and deaf customers who would otherwise not get through.

Other options apply to sign-in. According to the W3C examples, passkeys (WebAuthn), a sign-in link by email or a login through a third-party provider via OAuth meet 3.3.8, as do password fields that allow pasting and that password managers may fill. Recovering access also needs a route without a cognitive function test.

Which solution fits which form

Our assessment, without guarantee: contact and newsletter forms can often be protected with a honeypot, a server filter and a rate limit. For registration and password reset, an additional check in the background is worthwhile, and for sign-in what matters most is that pasting and autofill work. For high-risk flows such as payment, a honeypot alone is not enough, and developer sources point to dedicated solutions for defending against bots.

Questions to ask the provider

If you use a CAPTCHA or bot detection from outside, check the solution before you integrate it. Whether it removes barriers or creates new ones is something no marketing text tells you. Ask the provider these questions:

  • Second way: Does the solution offer a flow without typing, calculating and puzzles that is accessible to keyboard and screen reader?
  • Evidence: Is there an accessibility test report for the solution, and does it apply to the version running in your form?
  • Fallback: What happens with blocked cookies, disabled scripts or an ad blocker? The W3C draft describes cases with a fallback to classic CAPTCHAs.
  • Data: Which data does the provider collect, and where is it processed? This question belongs in the privacy review.
  • Support: How can customers reach a human when the protection rejects them, for example blind and visually impaired visitors?

How to check your CAPTCHA in five steps

Check accessibility by trying it out, with the same means your customers use.

  • Find every place: Contact form, newsletter, registration, login, forgotten password, checkout and reviews. Challenges that only appear after failed attempts or with an ad blocker count too.
  • Keyboard only: Operate the form with Tab, Shift+Tab, Enter and Space. Can you get in and out again without ending up in a keyboard trap?
  • With a screen reader: Test with NVDA (free) and, if possible, JAWS or VoiceOver. In the 2026 WebAIM survey, 55.0 percent name JAWS and 32.9 percent NVDA as their primary screen reader. Is the test announced, and does a path lead forward? If possible, have blind users test the flow. More in the glossary under screen reader.
  • A second way without typing: Answer the question of whether there is a way without typing, calculating and puzzles. Do pasting and password managers work in the password field and with codes?
  • Under changing conditions: Check in a private window, with blocked cookies, with an ad blocker and on a smartphone. 91.8 percent of respondents in the 2026 WebAIM survey also use a screen reader on mobile.

Document the results, because accessibility can only be demonstrated if it is clear what was tested. A single test does not replace a review by specialists. How such a test is set up is explained in our article Accessibility audit: scope, sampling and report.

SiteCockpit: keep your forms in view

Whether a puzzle can be solved by people is ultimately decided by a manual test. An automated scan, on the other hand, continuously checks many technical criteria, including forms, buttons and keyboard operation, and shows when changes to the shop introduce new errors. Digital accessibility is not a one-off project: themes, plugins and providers of spam protection keep changing.

  • easyMonitoring: Scans your pages regularly, including forms and keyboard operation, and delivers prioritized recommendations. History and comparison show whether your results are improving.
  • Manual audits: Specialists check what a scan cannot judge, for example whether a test can be solved by screen reader users.
  • Livecheck: The free quick test delivers a first assessment in under 60 seconds.

What makes sense now

No reason to panic, but real need for action: the standard is still in transition, yet a test without a second way already contradicts the rule from WCAG 2.1, and it costs customers. First clarify where your shop locks customers out.

  • Establish where you stand: List all CAPTCHAs and bot checks in the shop and work through the five checks.
  • Fix the most common errors: Replace tests that involve typing, calculating or puzzles with protection in the background, or offer an equivalent second way. Allow pasting in the password field and make sure every form offers a second way.
  • Keep it in view: After every theme or plugin update, check again whether the forms remain usable.

Proceed step by step and start with the forms through which customers order or sign in. With easyMonitoring you keep the technical criteria in view continuously, and for everything else an audit helps. Check every provider against the same criteria, and the accessibility of your sign-in will stay intact after updates and your forms will stay usable.

How accessible are your forms?

Test it for free with the SiteCockpit Livecheck and get a first assessment in under 60 seconds.

Test now for free

Table of Contents

View blog articles