ZeroSpams

CAPTCHA and accessibility: who gets locked out of your contact form

A CAPTCHA is built to be hard. The trouble is that "hard for a computer" turns out to mean "impossible for some people": customers who are blind, have low vision, tremors, dyslexia, or who are simply older or reading in a second language. For them, your contact form has a locked door.

Updated · 4 min read · By the ZeroSpams team

The short answer

W3C, the organisation that writes the web’s accessibility standards, states that every type of CAPTCHA will be unsolvable by users with certain disabilities [2], and that the interactive tasks CAPTCHAs rely on inherently exclude many disabled people [1]. The accessible answer for a contact form is protection that asks the visitor nothing: checks on your server that recognise bots instead of testing people.

What the accessibility standard says

W3C first published its note on the Inaccessibility of CAPTCHA in 2005 and updated it as a Working Group Note in 2019 [1]. Its central point: every interactive CAPTCHA asks the user to perform a task, and the very nature of that task excludes many people with disabilities, which amounts to denying them the service. The note also observes that many popular CAPTCHA techniques are no longer particularly effective or secure [1].

The Web Content Accessibility Guidelines (WCAG) go as far as saying CAPTCHAs deliberately push the edges of human ability, so that every type will be unsolvable for users with certain disabilities [2]. WCAG still permits CAPTCHAs, but only with a text description and an alternative form using a different sense, such as an audio version alongside an image. It permits them largely because banning them outright would push websites to ignore the guidelines altogether [2].

Who gets locked out

  • Blind and low-vision visitors: image grids and distorted text are unusable with a screen reader, and hard with magnification.
  • Deaf and hard-of-hearing visitors: the audio alternative doesn't help them, and the visual one may not either.
  • People with motor impairments or tremors: dragging a slider precisely, or tapping small image squares, can be impossible.
  • People with dyslexia or cognitive impairments: distorted letters and multi-step image tasks are exactly the kind of task that is hardest for them.
  • Older visitors: in the 2023 study of 1,400 people, solving time rose with age for all CAPTCHA types [4].
  • People reading in a second language: in Stanford's large study, participants outside the US took longer on English-language CAPTCHAs [3].

W3C also points to situational barriers that affect anyone: a small phone screen, bright sunlight, or a noisy place where audio can't be heard [1].

Why the audio option doesn’t fix it

Audio CAPTCHAs are the usual answer to "what about blind users?". The research is not kind to them. In Stanford's study of more than 1,100 people, three listeners agreed on an audio CAPTCHA's answer only 31% of the time, and each one took an average of 28.4 seconds [3]. And the audio version gives bots an easier way in: researchers defeated reCAPTCHA's audio challenge with off-the-shelf speech recognition [5].

So the accessibility option is at once the hardest version for people and one of the easiest for machines.

What about the "I’m not a robot" checkbox?

The checkbox is easier for most people than image grids, and W3C notes anecdotal evidence that it can be used with a range of assistive technologies. But when the background check isn't satisfied, it falls back to the same image or audio challenges [1]. Those fallbacks are more likely for people with unusual browsers or privacy settings, and assistive technology can itself make a browser look unusual. A visitor can't know in advance which version they will get.

Accessible ways to stop spam

W3C's note reviews approaches that don't ask the visitor to solve anything, including hidden "honeypot" fields and checks based on how the form is used [1]. For a contact form, the practical combination is:

  1. A hidden trap field, hidden correctly: out of view, skipped by the keyboard, and hidden from screen readers too, so no real person ever meets it.
  2. A minimum time before a form can be sent.
  3. A signed one-time token proving the form was started on your page.
  4. Link rules and a rate limit.

If you want one visible step, give it a keyboard and screen-reader alternative. Our slide-to-send control, for example, has a "press here to send" button that works with a keyboard and is announced correctly by screen readers. Nobody should need precise dragging to contact a business.

Why it matters for your business

Beyond fairness, it is simple arithmetic: every customer who can't get through your form is an enquiry lost, and they rarely tell you. Accessibility is also increasingly a legal expectation for business websites in many countries, so an inaccessible form can be a compliance risk as well as a lost sale. Check what applies where you operate.

The wider research on lost customers is in is CAPTCHA costing you customers?

Common questions

Is reCAPTCHA accessible?

Partly. The checkbox itself can work with assistive technology, but when Google’s background check is unsure it falls back to image or audio challenges, which many disabled people can’t complete [1].

Does WCAG forbid CAPTCHA?

No. WCAG allows CAPTCHAs if they have a text description and an alternative form using another sense. It also acknowledges that every type will be unsolvable by some users with disabilities [2].

Is a honeypot field accessible?

Yes, if it is hidden properly: positioned out of view, removed from keyboard navigation and hidden from screen readers, with a label telling anyone who does find it to leave it empty. Then no real person ever encounters it.

Is a slider CAPTCHA accessible?

Not on its own: dragging can be hard or impossible with a tremor, a switch device or a screen reader. A slider needs a keyboard-accessible button that does the same thing.

Sources

Numbers in this guide come from these studies and publications. Links open the original.

  1. Inaccessibility of CAPTCHA: Alternatives to Visual Turing Tests on the WebW3C Accessible Platform Architectures Working Group (Working Group Note 2019, with later draft updates)
  2. Understanding Success Criterion 1.1.1: Non-text Content (CAPTCHA)W3C Web Accessibility Initiative (WCAG 2.1)
  3. How Good Are Humans at Solving CAPTCHAs? A Large Scale EvaluationBursztein, Bethard, Fabry, Mitchell, Jurafsky (Stanford University). IEEE Symposium on Security and Privacy, 2010More than 1,100 participants, 318,000 CAPTCHAs.
  4. An Empirical Study & Evaluation of Modern CAPTCHAsSearles, Nakatsuka, Ozturk, Paverd, Tsudik, Enkoji (UC Irvine, ETH Zürich, Microsoft). USENIX Security Symposium, 20231,400 participants solved 14,000 CAPTCHAs; 200 popular websites inspected. Participants were paid online workers.
  5. Dazed & Confused: A Large-Scale Real-World User Study of reCAPTCHAv2Searles, Prapty, Tsudik (UC Irvine), 202313 months, more than 3,600 real users who did not know they were being studied.