Does Accessibe Work?

Or is automated accessibility a scam?

Over the past several years, a new class of accessibility software has taken the web by storm, claiming that it can make a website 100% accessible without any further effort by the site owner. Accessibe is the most prominent of these entrants, but there are a number of other competitors offering a similar value proposition. The concept is obviously very appealing; who wouldn’t want to make it easier for sites to become accessible?

Site owners are happy to avoid expensive audits, remediations or rebuilds without being vulnerable to lawsuits under the ADA. And venture capital has made multimillion dollar investments in these services in the expectation that it’s a growth market.

The question: Is it too good to be true?

Unfortunately, the answer is yes. While automated accessibility services can successfully address some common accessibility issues, they are not capable of reliably making a website 100% accessible without any violations of WCAG 2.1 AA as claimed. I have spoken to representatives of several of these companies and the honest ones will acknowledge the limits of their software if asked. But even the companies that are less forthcoming will give you a clue if you read their terms and conditions; suddenly the claims they make sound a lot less ambitious.

What’s the harm in using an accessibility overlay?

Within the accessibility world, these types of automated solutions are referred to as “overlays”. This is because they are software that is overlaid on top of the existing website; the service makes no changes to the underlying site code or functionality but acts as a layer that modifies or translates the site content in the browser when a visitor interacts with it. Let’s stipulate that automated accessibility overlays can address some (but not all) accessibility issues; while it won’t solve all of your problems, is there any harm in using one?

Accessibe can degrade the user experience

Automated accessibility overlays interfere with screen readers

Screen reader users and other members of the accessibility-focused community have criticized Accessibe for hijacking the user experience such that it interferes with the way that they normally use JAWS, NVDA and other screen readers. In fact, there is now a browser plugin that can be used to block Accessibe and other overlays from loading so that screen reader users can browse in peace. Accessibe, as the most prominent and successful service in the automated accessibility space, has been a lightning rod for the harshest criticism, but many of these points apply more broadly to the other players in this market.

Net-new accessibility violations

This isn’t only a matter of a poor user experience. The Accessibe widget itself contains several accessibility violations. One easy to understand example is the language picker that allows users to select from multiple languages that the service supports. The button shows your current (or default) language (which is English in my case). When screen readers focus on that button, it does not announce the purpose of the button (e.g. “Select an interface language”) but only announces the current language name (e.g. “English button”).

Even when opening the language selection popup, the interface is not itself accessible. When each language is announced in a localized manner (e.g. Espanol for Spanish), the screen reader does not announce that the language has switched temporarily from English to Spanish. While this is perhaps a minor issue when it comes to languages that use Latin characters, for languages like Hebrew, Chinese, or Russian, the language announcement would be unintelligible to a native speaker.

Automated accessibility is frequently redundant

Furthermore, while most of the automated accessibility services claim to alter the website’s code output to make it more accessible, these products are not invisible. All of these services’ widgets include an interactive set of controls that allow users to manually make adjustments to text size, colors, and other characteristics in order to fit their own browsing preferences.

The complaint here is similar. Users with disabilities already have tools, many of them already built into browsers, to modify default contrasts and text size as needed. At best, these interfaces are irrelevant window dressing that appear to make the website more accessible than it is and at worst they can get in the way when users with disabilities attempt to make use of their existing tools.

This is all true; the most visible portions of popular accessibility overlays are largely redundant with tools that are available within any browser for free. Nonetheless, I do want to give credit where due that I think these automated accessibility services are onto something here. While it’s common in the accessibility world to note that nearly 20% of United States residents have some sort of disability (per the U.S. Census Bureau), I think it is reasonable to assume that many of these people are unaware of the accessibility-friendly tools that are built-in to browsers. Speaking from personal experience, many older relatives who would benefit from assistive technology have little idea of their existing browsing features nor are they inclined to learn how to use them. Building these features directly into the web interface may be a better solution for these users (as long as there is a way for power users to turn it off).

Does it succeed on its own terms?

Automated accessibility can only address some types of issues

We’ve established that screen reader users and other power users in the accessibility world dislike automated accessibility overlays. That alone is a good reason to avoid using such a tool. But even if you are only looking to address accessibility purely from a legal compliance perspective, will services like Accessibe do the trick?

A quick perusal of Accessibe’s terms of services (as of November 14, 2021) can reveal some of the inherent limitations of the service.

• Most media and files including audio, video, SVG images, and PDFs are not covered. This excludes important WCAG requirements such as closed captioning for videos among others. Site owners who require help making these assets accessible are invited to pay for additional (presumably manual) services for an additional fee.
• The service is based on artificial intelligence (AI) and as a result, cannot necessarily automatically remedy unique or uncommon functionality.
• Should you notify Accessibe of any WCAG violations that you find, they agree to assist you only insofar as these issues are specifically introduced by Accessibe.
• Prior to installing Accessibe, the customer must verify that the site contains no HTML validation errors (among other things). Believe it or not, lack of HTML validation errors is actually one of the accessibility requirements in WCAG (4.1). One more item that the service does not even attempt to remedy.
• It is also the customer’s responsibility to test and verify the functionality of Accessibe’s solution on their own website.

Adrian Roselli has noted a number of other troubling disclaimers and limitations embedded in earlier versions of the terms (as of February 2021); readers can draw their own conclusions as to whether the latest version of the terms represents improvements in the services or merely more careful drafting from Accessibe’s legal and marketing teams.

Beware of AI claims

Accessibe in particular markets itself as being powered by AI, which allows it to automatically understand the meaning and purpose of site elements and adjust them accordingly to make them accessible. We are not artificial intelligence experts by any means but consider us skeptical. Most of the examples touted on Accessibe’s feature page describing its AI refer to items that could be relatively easily programmed without any machine learning or AI (such as identifying buttons, form fields, or menus).

The one feature that clearly does rely on some form of artificial intelligence is image recognition. (It’s unclear whether this capability is homegrown or licensed; Google, Amazon, Microsoft and other prominent tech companies license access to their image recognition services at very reasonable rates.) There are some common use cases where image recognition can improve accessibility; for example, identifying 5 stars as a rating or a Facebook icon as a link to a social media page. 

However, in our opinion, image recognition can cause more confusion than it clears up (even when it works as intended). One common accessibility violation is adding alternative text for decorative images; those images do not actually contribute to the user's understanding of the page and should generally not have alternative text. Or let’s say an e-commerce site shows a photo of a model frolicking in a meadow wearing a dress; Accessible’s AI might emphasize the meadow rather than the product being sold. There is no way to be sure that the image recognition is describing the correct thing in context.

But let’s set these reservations aside for now. Remember how we noted above that many of the accessibility features touted by automated accessibility services are already available within browser settings? Would you be surprised to learn that Google Chrome offers access to automatic image descriptions for free (in English, French, German, Hindi, Italian, and Spanish)?

Can Accessibe protect you from lawsuits?

No.

Sites using Accessibe have been the targets of a number of lawsuits. It’s true that nothing out there can prevent someone from filing suit against you (even if it’s extremely frivolous). But you would expect the plaintiff lawyers to be wary of guaranteed loss if Accessibe were as effective as it claims to be.

If you read the terms of service that I highlighted above, you’ll remember that there are quite a few things that Accessibe says that you are responsible for taking care of before its service can work. Beyond that, Accessible does offer a litigation support package in which they pledge to provide you with information that may be useful in defending against a lawsuit (if you are sued). That’s all well and good, but if push comes to shove, you are on your own. The boilerplate language in the terms specifically notes that they make no warranties and assume no liabilities. I am not a lawyer and I don’t know whether these terms can hold up in court, but would you want to take your chances?

It is much more troubling that at least one of these lawsuits (Murphy v.  Eyebobs) references net-new accessibility issues that are present in Accessibe’s own overlay (aside from documenting instances where the overlay fails to correct existing accessibility issues that are present on a website). If this is the case, it means that Accessibe that not only fails to bring websites into compliance with the ADA, but it actually may do more harm than good.

This past month, Eyebobs agreed to a settlement that requires them to hire an accessibility consultant to manually address its website’s accessibility. This story in WIRED shares more details about the settlement including this shocking quote:

So much for the litigation support package.

If these issues are not corrected, it would not be surprising if smart and savvy lawyers started treating the presence of an Accessibe overlay as a sign that they will easily find accessibility violations. And if Accessibe increases your legal risk (instead of eliminating it), is it even worth $50/month?

The Tragedy of Accessibe: Right Diagnosis. Wrong Prescription.

The tragedy of this whole situation is that we think Accessibe’s critique of the larger digital accessibility market is spot on. For most businesses, finding and fixing their websites’ accessibility issues is far too expensive. This is especially the case when you consider how much the average company spends on their website. It would be not be an exaggeration to say that the typical accessibility audit and remediation project costs more than the average web project.

But even if you set aside the issue of cost, manual accessibility has a hard time scaling. Even if every single site owner in the world decided tomorrow to make accessibility a priority, who would they hire? It would take many years for the existing set of accessibility professionals to address this problem.

And finally, most businesses approach the question of accessibility from the perspective of legal risk. We don’t think that’s ideal, but it’s a fact that isn’t going to change any time soon. (In fact, even getting business owners to understand that inaccessibility can introduce legal liability is a huge victory for disability rights activists and legal advocates.)

If the current market is not accessible to most websites, it’s to be expected that they are receptive to alternatives like Accessibe that claim to address that legal risk at a very attractive price point. We always like to say that more accessible is better than less accessible and if Accessibe and other overlay vendors were more forthcoming about their actual capabilities, I think they could be a net positive for the world (especially if there were a way for screen reader users who find that it interferes with their existing assistive technology to opt out entirely).

The response from the established accessibility industry to these very compelling points is underwhelming. To their credit, the community produces tons of free resources and documentation that help to introduce accessibility concepts to site owners, content managers, and developers but this is not very helpful to the average business website. But beyond that, the rhetoric that I hear is that each company has a moral obligation to try harder and that the simplest way to maintain accessibility is to build a site the right way the first time. All true, but not particularly helpful to those without the technological know-how.

At Access Armada, we definitely don’t have all the answers, but we are attempting to find ways to dramatically improve accessibility for small to medium size businesses at an accessible price point. If you’re interested in learning more, please reach out!