Accessibility overlays: the documented record.

What an overlay promises, and what it does

An overlay is a third-party script you paste into your site. It adds a floating accessibility button offering font sizing, color modes and a “screen reader” profile, and the marketing promises that AI will detect and repair accessibility problems automatically in the browser. One snippet, instant compliance, no code changes. For an owner under pressure that is an almost perfect pitch.

Start with the part that is true, because a page that says overlays fix nothing is easy to disprove and we would deserve it. The Overlay Factsheet, the practitioner consensus document signed by people who write the specifications, says a non-trivial array of accessibility problems genuinely can be repaired this way. Toolbars that enlarge type or raise contrast do help some people. The same document then names exactly where the repair stops being reliable, and the list is the load-bearing part of any site. Text alternatives for images, form labels and error handling, focus control, keyboard access, and anything a React, Angular or Vue application changes underneath the script. Whole formats go untouched, including PDF, canvas, SVG and media.

So the argument is not that the technology does nothing. It is that its conclusion is the one the Factsheet reaches, which is that full compliance cannot be achieved with an overlay and legal risk therefore cannot be eliminated by one. The rest of this page is what happened when the stronger pitch met blind users, courts and regulators.

What blind people's own organizations said about it

The clearest voice belongs to the National Federation of the Blind, the largest organization of blind people in the United States. In June 2021, its board took the extraordinary step of revoking accessiBe's sponsorship of the NFB national convention, stating publicly that the company engaged in behavior “harmful to the advancement of blind people in society” and that it dismissed the concerns blind users raised about the product.

Weeks later, at its July 2021 national convention, the membership passed Resolution 2021-04, which condemns the marketing of automated overlays as a category and names six vendors in its opening clause. In the NFB's words, claims of achieving WCAG compliance “through the integration of one line of code” are false. A second resolution from the same convention, 2021-17, deals with accessiBe alone, and both numbers circulate, so it is worth knowing which is which.

Read 2021-04 carefully and it is more interesting than a condemnation, because it also sets out what an overlay would have to let somebody do to be worth having. Know that it is there. Know when it is running. Switch it on and off. Configure it with controls that are themselves accessible. Reach whoever built it. That is an attack on how the category is sold rather than on the idea that software can help, and it is a distinction a vendor's lawyer will draw if we do not draw it first. The resolution's own conclusion is that overlays may help make certain parts of a website nonvisually accessible and alone cannot guarantee full nonvisual access.

What blind users report in practice runs alongside that. The widgets interfere with the screen readers and keyboard setups people already have, trap focus, override working preferences, and sit on top of the same broken code. Which points at the quieter problem underneath the whole category. Somebody who needs magnification needs it on every site they visit, so they already have it, configured years ago, working everywhere. A widget offering them a worse copy of their own software on one site is solving a problem they solved before they arrived.

The Eyebobs case: overlay installed, sued anyway

In January 2021, a blind screen reader user sued the eyewear retailer Eyebobs in federal court in western Pennsylvania. Eyebobs was running accessiBe at the time, and the complaint did not treat that as a defense. It treated it as evidence, pleading that the overlay failed to give screen reader users full and equal access to the platform. Accessibility consultant Karl Groves filed a declaration for the plaintiff setting out why the case should proceed despite the widget.

It settled, and the settlement asked for what the widget had promised to make unnecessary. Conformance work, an accessibility consultant, an internal accessibility team, staff training, and annual testing by end users. Be careful with the dates you will find quoted for it, ours included. A settlement of this shape gets preliminary approval first and final approval later, after a fairness hearing, and plenty of write-ups collapse the two into one. What is not in dispute is the shape of the deal.

A second settlement put the point in writing that December, when the National Federation of the Blind and blind employees resolved a case against ADP, which had been running an AudioEye overlay. The agreement records the parties' position that overlay solutions of the kind then supplied by companies such as AudioEye and accessiBe do not suffice to achieve accessibility. That is negotiated language rather than a judge's finding, it speaks to the products as they then were, and it binds nobody outside that case. Two settlements, one conclusion, both in the public record.

What happened when the overlay defense reached a judge

Twice in 2021, businesses running accessiBe asked federal courts in New York to dismiss accessibility cases against them on the strength of the widget. Both times the court refused. The cases are Angeles v. Grace Products and Quezada v. U.S. Wings, both in the Southern District of New York, and both are worth naming because vague references to what courts have said about overlays are exactly the problem. In one of them the defendant had bought a year-long partnership with the vendor and had been given three audit reports concluding the site complied with WCAG 2.1. The court's answer was that the defendant had not shown the site's accessibility problems were fixed beyond doubt, because barriers were still alleged to exist. The second decision turned on the same reasoning, that a defendant carries a formidable burden to show a site has been brought into compliance and will stay there.

If somebody has handed you a compliance certificate from an overlay vendor, that first case is the one to read. It is the occasion when exactly that document was put in front of a judge, and it did not carry the defense.

Now be clear about what those decisions are not, because plenty of pages on this subject are not. Both were rulings on motions rather than trials, and both turned on the defendant failing to carry a heavy burden of showing the case was over, not on any finding about the widget itself. The two settlements above were negotiated rather than decided. So we have found no decision anywhere holding that an overlay is or is not adequate under the ADA, and if somebody shows you one, read it before you rely on it. What courts have declined to do, twice, is treat a widget as proof that the problem is solved.

The practitioners are less equivocal. In January 2021 WebAIM asked accessibility professionals to rate these tools, and 676 of them answered that question. Just over two thirds put them at not very or not at all effective: WebAIM's own summary calls it 67%, and adding the two published percentages in its own table gets you 69. It rises to 72% among respondents who have disabilities themselves. Worth naming the other end of it, since 31% rated them somewhat or very effective, and quoting only the first number would be the selective citation we criticize vendors for. WebAIM also says its sample was not controlled and may not represent practitioners generally, so read it as what the people doing this work think, not as a measurement of the products.

One observation of ours to close, offered as ours. A third-party script sits in your served HTML where anyone can read it, which means a widget is visible to a plaintiff's firm before they open a single page. What it tells them is that somebody spent money on the problem and that the code underneath is probably unchanged, because changing it is the one thing the script does not do.

What the FTC alleged, and what its order requires

In January 2025 the Federal Trade Commission accepted a proposed consent order under which accessiBe would pay the FTC $1 million that may be used to provide refunds to consumers, over advertising that said the widget could make any website WCAG compliant. The Commission approved the final consent order on a 3-0 vote in April 2025, and it runs for 20 years from issuance, with a sworn compliance report due at one year. The 20 years can run longer, because the order restarts that clock from any later complaint filed over a breach of it. Two things in that sentence get lost constantly, and both matter. It is a consent order and not a fine, and the company neither admitted nor denied the allegations, so nothing here is a court's finding of fraud.

The order itself is the durable part. It bars accessiBe from representing that its automated products can make any website WCAG-compliant, or keep it compliant as the content changes, unless the representation is not misleading and the company possesses and relies on competent and reliable evidence for it at the time it makes the claim. That last clause is the one with teeth, because it puts the evidence in front of the advertisement rather than behind it. A separate provision bars misrepresenting reviews and endorsements as independent, which is narrower than a ban on paid coverage and is the thing the order actually says. And it requires the company to disclose, before a customer takes on any financial obligation, that the product will not correct barriers on third-party domains that are not themselves running it. Read that last one against your own site. If your checkout, your booking step, or your payment page is hosted somewhere else, a federal regulator has already decided that the gap needs disclosing at the point of sale.

The strongest sentence in the whole case is not in the order at all. It is in the complaint, at paragraph 78, where the Commission states that according to web accessibility experts, including the W3C and the developers of the WAVE evaluation tool, no automated testing tool alone can determine if a website meets accessibility standards, and manual human testing is required. That is a US federal agency, in a filed pleading, stating the argument this whole business rests on. It is what the FTC alleged rather than what a judge decided, and it is a long way from marketing copy. The National Federation of the Blind filed formal comments supporting the consent order as written.

Why the widget was recommended to you

Three things in the record explain how these scripts reach sites whose owners never went looking for one. Take them in order, because they stack.

The first is a commission. The FTC's complaint records that accessiBe offers agency partners a twenty percent discount on its standard pricing, which the agency may pass on to the client or keep for itself. Nobody has called that unlawful and it is an ordinary reseller arrangement. It is also the mechanical answer to a question owners put to us often, which is why their web agency was so enthusiastic about a widget.

The second is the reviews, and roughly half the FTC's case was about those rather than about the technology. The complaint puts $1,900 behind one published review of the product and $850 behind another, with the widget installed free on the publisher's own site in both cases. It also states that the company hired a marketing firm to generate reviews designed to look as though they came from independent publications, and to move Google's results. The final order bars misrepresenting third-party reviews as the independent opinion of impartial authors. Which means the glowing write-up you found while researching this may have been part of what you were researching.

The third has nothing to do with marketing. Monitoring platforms now sell overlays as modules. Acquia's Web Governance, the product most buyers still call Monsido, is a crawler that reports on your site, and two paid add-ons change it instead. PageCorrect applies fixes to the live page through the platform's script. PageAssist puts an accessibility toolbar in front of your visitors, and Acquia's own description is that it makes changes to your site in real time. Those two together are the halves of an overlay, arriving as options on a reporting tool. So an organization whose procurement policy rules overlays out can end up running one because another department bought a dashboard. Which switches turn it on is worth checking before renewal.

What Germany's accessibility office publishes

With the European Accessibility Act in force, the overlay question landed on regulators' desks, and Germany's federal accessibility office, the Bundesfachstelle Barrierefreiheit, answered it in writing. Its guidance for online shops lists “no use of overlay tools” as one of the numbered steps, on the ground that an overlay cannot make a web presence fully accessible and that an online shop cannot meet its BFSG obligations with one.

The same paragraph goes further than anything else on this page, and it comes from a government body rather than from a competitor. It records that using such tools often creates further barriers in a website that would not have existed without the tool at all. The office is fair about the other side too, and says plainly what an overlay can do, which is let a visitor adapt the appearance to their own needs with larger type and a different background color.

Now the qualification that most write-ups leave off, and it matters. The Bundesfachstelle's statutory job is information and advice. It is not the enforcement authority, and its own BFSG questions and answers point at the states' joint market surveillance body, the MLBF, for that. So this is published guidance from a national advisory office, not a regulator's decision in a case, not a ban, and not a rule that applies across the EU. It is still the best documented answer any European government has put in writing to the question you are asking.

Even the customers are suing now

The latest turn is overlay vendors being sued by their own customers. In the summer of 2024 an online florist filed a class action against its overlay provider in federal court in Delaware, after buying the widget and getting sued anyway. It is at least the second such case documented by disability rights lawyer Lainey Feingold, who has tracked overlay litigation for years. In February 2026 a magistrate judge recommended the consumer fraud and negligent misrepresentation claims go forward. When a product's own customers become its plaintiffs, the pattern is complete.

The vendors have gone to law against their critics

Most pages about overlays leave this out, and it belongs on one, because it changes what you are able to find out before you buy.

In France, the overlay company FACIL'iti sued Julie Moynat, a web developer, in her personal capacity, over what she had written about overlays. The subpoena asked the court to find denigration, to award damages, and to order her to stop writing anything that discredited the product on any medium at all, naming social networks, blogs, websites and webinars. The same company also pursued Koena, a seven-person French accessibility consultancy. In late November 2023 a judgment ordered Koena to pay FACIL'iti €26,256 and to delete two tweets. Koena was appealing, and we do not know how the appeal came out.

In the United States, AudioEye sent a cease and desist letter in April 2022 to Adrian Roselli, an accessibility expert who had contributed to the Overlay Factsheet. He published the letter with his reply, and the line worth carrying is his own. Threatening genuine criticism with expensive legal action, he wrote, is not in the best interests of the people both sides claim to want to help. No lawsuit followed. A letter from a company's lawyers is usually enough on its own.

Here is why that matters to you rather than to them, and this last step is our read rather than the record. When you go looking for an honest assessment of a product in this category, some of the people best placed to write one have been sued or threatened for doing exactly that. A thin public record is a reason to lean harder on two things that cannot be chilled, which are what a regulator has published and what a vendor has published about its own limits. Both of those are on this page.

Why the technology cannot close the gap

• Your source code is untouched. A div pretending to be a button, an unlabeled form field, a checkout that cannot be finished by ear. The overlay paints over them as the page loads and often guesses wrong, and the defects a complaint would cite are still sitting in the HTML for any expert to find.
• Automation only reaches the part a machine can settle. We graded 356 of the 432 techniques and failures we publish, and 10 can be settled by a machine outright. That grading is ours rather than the standard's. Be precise about why the rest is hard, because the loose version of this argument is wrong. A script can change the page and the accessibility tree, and W3C publishes a technique that does exactly that. What no script can do is know things it was never told: which of two products a photograph is selling, what your error message should advise, or which order a screen is meant to be read in. Guessing those is the part that goes wrong, and it goes wrong silently.
• Stop paying and the problem comes straight back. One vendor's own documentation states it plainly, that when the subscription ends the error remains in the website's HTML, and recommends developers fix the issue in the code instead. A rented fix is a fix you are renting.
• It cannot fix what it cannot see. Documents of every kind, audio and video, canvas, SVG, and anything inside a third-party frame are all outside its reach, and so is any domain other than yours. For most stores the checkout, the booking step or the payment page is exactly one of those things.

What an overlay works out about your visitors

An overlay that switches settings on by itself has to detect that assistive technology is running, and detecting that means working out that the person at the keyboard has a disability. Information about someone's health or disability is sensitive personal data almost everywhere it is regulated. That is the whole argument and it takes two sentences. What follows from it takes longer.

Some of these products remember the settings from one site to the next, using a cookie, with nobody having opted in and no obvious way to opt out. Somebody who turned on a screen reader profile at one shop can arrive at the next one already flagged, by a script neither site's owner wrote.

Now the part that decides whose problem this is. The Overlay Factsheet, the same document we cite above for what overlays genuinely repair, says the exposure lands on the overlay customer. That is you. The script runs on your domain, under your privacy notice, with your company name at the foot of the page, and a visitor who objects writes to you rather than to whoever wrote the code. If GDPR or California's privacy law reaches your business, that is a second reason to know what the widget is doing, completely separate from the accessibility one and landing on a different person in your building.

None of this says the detection is done in bad faith, and none of it is a legal conclusion about any particular product. What a given widget collects, whether it identifies anyone, and what it infers are questions you answer by inspecting the thing on your own site, which we have not done for you. Responsibility can land on the vendor as well as on you, depending on who decides what and who handles what, so this is a conversation to have with both your data protection adviser and the vendor rather than a bill you have already been handed. The point is narrower and it still stands. Most owners pasted the script in without being told there might be any processing involved at all.

So… is it a scam?

It is the word people actually type into a search bar, so it deserves a careful answer rather than a cheer. “Scam” implies intentional fraud, no court or regulator has made that finding, and we are not going to make it either. The record is harsh enough without it. The FTC's own release described the claims it was acting on as deceptive and the Commission ordered a $1 million payment. The NFB called the flagship vendor's conduct harmful to blind people. A settlement agreement in a different case said overlays will not suffice. Two federal judges declined to treat one as proof a site was fixed.

Whatever word you land on for that pattern, the practical conclusion does not move. Money spent on an overlay buys some genuine repair, and it does not buy conformance and it does not buy legal safety, which are the two things it is sold for. When a product's core promise fails the people it names, gets refused as a defense, and draws a federal deception order, you do not need the word “scam” to know what to do next.

Sources: every claim above, in the public record

• NFB: convention sponsorship statement on accessiBe (2021)
• NFB: 2021 resolutions (2021-04 on overlays)
• FTC: final order requiring accessiBe to pay $1M
• NFB: comments to the FTC on the accessiBe consent order
• Equal Entry: the Eyebobs lawsuit, explained
• Law Office of Lainey Feingold: overlay legal updates
• Feingold: small business sues its overlay vendor (2025)
• Bundesfachstelle Barrierefreiheit: BFSG and online shops (Germany)
• Overlay Factsheet: practitioner consensus document
• WebAIM: the January 2021 practitioner survey, with the overlay table
• FTC: the decision and order itself, in the case file
• Adrian Roselli: the cease and desist letter he received from AudioEye
• Acquia: PageAssist, the visitor toolbar sold as an add-on to a monitoring platform

Facts current as of August 2026. Everything said here about a named company comes from the linked public record, meaning court filings, agency orders, and what organizations have published themselves. Where the record is an allegation rather than a finding, the page says so in the sentence.