What is AudioEye is the easiest question in this category to answer honestly, because AudioEye is the only vendor of its kind that has to file quarterly.
It trades on Nasdaq as AEYE. That means its revenue, its customer count, its channel mix and its losses are public, audited and checkable, while Siteimprove, Silktide, Level Access and Deque publish only what they choose to.
Use that. Divide the numbers and the company explains itself.
AudioEye was founded in 2005 and is headquartered in Tucson, Arizona. As of its second quarter of 2026 it reported $10.7 million in quarterly revenue, $42.3 million in annual recurring revenue, and roughly 129,000 customers. Kelly Georgevich became chief executive in May 2026, moving up from chief financial officer.
Divide $42.3 million by 129,000 customers and you get $327.91 per customer per year. About $27 a month.
Hold that number. Everything else in what is AudioEye, as a business and as a product, follows from it.
This piece covers what the product actually does at runtime, what the filings reveal that the marketing does not, the overlay question reported carefully, the developer tooling that is not an overlay at all, who it genuinely fits, and where the critics themselves concede ground.
The short version: AudioEye is two businesses wearing one brand. A high-volume, low-ticket, script-injection marketplace business that is growing, and a flat enterprise services business that does the expert work the marketing describes. The marketing collapses them. The quarterly filing separates them.
What is AudioEye doing at runtime
AudioEye describes itself as combining "industry-leading AI automation technology with expert fixes informed by the disability community."
Its own technical explainer, dated July 2026, is admirably clear about the mechanism:
"When a page loads, AudioEye’s JavaScript reads the rendered code, identifies accessibility issues, and fixes them before the user ever interacts with the page."
The named operations are adding missing labels, correcting heading structure, improving ARIA implementation, fixing form associations, clarifying links and exposing navigation structure. The company states that automated fixes resolve 50 percent of issues in real time, and that automation tests 37 of the 55 WCAG 2.2 Level A and AA criteria.
And here is the single most important sentence the company publishes, from that same explainer:
"when our experts identify an issue, the fix doesn’t just live in a report. It deploys through the same JavaScript framework, meaning human insight becomes automated protection applied at scale across every page."
Read that carefully. The human labor is real. AudioEye employs auditors and acquired two audit businesses, Bureau of Internet Accessibility in 2022 and ADA Site Compliance in September 2024. But the delivery mechanism for expert remediation is still the injected script, not a commit to your source code. Stop paying and the work leaves with the subscription.
AudioEye draws its own line against toolbar products, and it is a real distinction worth understanding: a widget "changes what a sighted user sees. AudioEye changes what assistive technology reads."
What the filings say that the marketing does not
Three things, all from AudioEye’s own reporting.
The customer count is not growing. The sequence runs roughly 127,000 in December 2024, 120,000 in June 2025, 123,000 in September 2025, 131,000 in December 2025, and 129,000 in June 2026. The base fell in the first half of 2025 and again in the first half of 2026. The headline "up 9,000 year over year" is measured from the trough. Annual recurring revenue rises while the customer count shrinks, which means growth is coming from price and mix rather than from new logos.
The business is two businesses. In the second quarter of 2026, Partner and Marketplace revenue was $6.24 million, up 16 percent year over year. Enterprise revenue was $4.47 million and roughly flat. So 58 percent of revenue and essentially all of the growth comes from the channel where the product is a script you paste into a hosted CMS.
The GAAP loss is widening while the adjusted number sets records. Adjusted EBITDA hit a record $3.0 million in the quarter at a 28 percent margin. The GAAP net loss for the first half of 2026 was $2.98 million against $1.47 million a year earlier, roughly doubling, with general and administrative expense up 30 percent on litigation, severance and amortization.
And the arithmetic that ties it together. AudioEye reports around 116 employees against roughly 129,000 customers. That is over a thousand customers per employee. At $328 a year blended, the human expert layer cannot possibly be reaching the general customer base. It is concentrated in the enterprise 42 percent of revenue, which is the half that is not growing.
What it costs
AudioEye no longer publishes prices. Its plans page lists three tiers, Automated, Self-Managed and Managed, described in words with no dollar figures and a mix of "Get Started" and "Talk to an Expert" buttons. The pricing and plans URLs return 404.
Third-party sources conflict wildly. One 2026 review lists $199, $399 and $799 a month. Another lists $39, $99 and $599. They cannot both be current, and neither should be quoted.
Use the derived figure instead. $327.91 per customer per year, from the company’s own filings. That number is honest, it is checkable, and it sits two orders of magnitude below the top tier those reviewers quote, which tells you exactly where the customer mass actually is.
There is a reseller and platform partner channel offering wholesale rates and bulk discounts, with named relationships including Duda. No revenue share or wholesale rate is published. That channel plus marketplaces is the majority of revenue and essentially all of the growth.
The overlay question
AudioEye is named on the Overlay Fact Sheet, which carries more than a thousand signatories and states that no overlay product can make a website fully compliant with any accessibility standard.
AudioEye’s rebuttal is more interesting than a flat denial. It splits the term. A general overlay is "a javascript-based technology that sits on top of website code to provide an added layer of functionality," and it points out that Google Analytics fits that description. An overlay widget is "an onsite widget or toolbar meant to fix accessibility issues." It then rejects the second category outright:
"Overlay widgets are not part of the solution. They have no business masquerading as an accessibility solution."
That is a genuine position, and it separates AudioEye from vendors that sell a toolbar as a compliance product. Worth noting that the US Federal Trade Commission’s $1 million order in January 2025 over claims that an automated product could make sites compliant was against accessiBe, not AudioEye.
On the study you may have seen. In February 2026 a press release announced that an "independent study" found AudioEye’s detection exceeded the industry by 89 to 253 percent. The press release does not disclose who paid for it and does not name the competing tools.
The landing page does both. It states plainly that "This study was commissioned by AudioEye and conducted independently by Adience," and names the five tools tested as AudioEye, accessiBe, Deque, EqualWeb and UserWay, across six websites.
So the disclosure exists. It is just placed where far fewer people will see it than the wire story that syndicated everywhere. That is the criticism worth making, and it is narrower and fairer than "they hid it." Two further limits: the study measures detection count only, not fix quality or user outcome, and more issues flagged is not the same as more issues fixed when false positives inflate counts.
We found no independent, non-commissioned, published controlled test of AudioEye’s remediation effectiveness in 2025 or 2026. That absence is itself the finding. The critical technical literature is argument and demonstration. The quantitative material is vendor-commissioned.
One documented product-performance fact, reported as fact and not as legal analysis: in a 2021 settlement involving a company that was using AudioEye, the agreement stated that overlay solutions such as those provided by AudioEye and accessiBe "will not suffice to achieve Accessibility." Separately, an industry tracker reports that lawsuits against sites already running accessibility widgets have continued to rise, and that the presence of a widget "has not proven to be a meaningful deterrent, legally or practically." Those figures cover all widget vendors, not AudioEye specifically.
There is also a matter of record worth stating once. AudioEye sued the accessibility practitioner Adrian Roselli for defamation in 2022 and dismissed the suit with prejudice in January 2024, with settlement language acknowledging that the statements at issue "were merely expressions of Roselli’s opinion," and a contribution of at least $10,000 to the National Federation of the Blind. It is the most damaging documented episode in the company’s relationship with the practitioner community.
The part that is not an overlay at all
This gets missed, and it is the most useful section for a developer audience.
AudioEye ships a Testing SDK and CLI that tests "a reachable URL, including internal, staging, or deployed pages," a local HTML file, raw HTML over standard input, or a component fragment, with HTML, JSON and CSV output. The developer tooling page describes testing in pre-production environments, CI pipeline installation to scan on every deployment, Jest for components, Cypress and Playwright for multi-page flows, and CSV export to Jira or Asana.
That layer produces source-level findings for developer remediation. It is not an overlay. Its documented limitation is stated plainly: "The CLI is best for quick scans of a page URL or local HTML. It does not support cookies or authenticated flows." Authenticated flows require the Playwright or Cypress integrations.
For a higher-education or government reader with developer capacity, the Self-Managed tier and the SDK are the defensible product here, and the runtime script is not.
On the production side, installation into WordPress, Shopify, Squarespace, Wix, Drupal, Duda, BigCommerce, Webflow, Joomla, HubSpot and others is a script paste into the footer. There is also a Finalsite integration, which matters if you work with schools.
Who it genuinely fits, including the critics’ own concession
The strongest case for an overlay comes from the person who maintains the Overlay Fact Sheet. Karl Groves has written that overlays "can offer a quick fix" when "accessibility issues are egregious and the organization is under considerable outside pressure," with the conditions that it is temporary relief and that the organization must in parallel plan real remediation and train its developers.
His cost objection is the one to carry into a client conversation:
"any skilled JavaScript developer with good accessibility knowledge can do the same thing, without the need to pay for a third-party service"
And his technical objection: "This fixes the symptom without treating the disease," and if the underlying interface code changes, the fix breaks.
So what is AudioEye actually for? Genuine fit: a small nonprofit or association on a hosted CMS with no developer, no build pipeline and no audit budget, for whom roughly $27 a month buys automated detection plus real runtime repair that would otherwise not exist at all, while a remediation plan gets funded.
Genuine misfit: any organization with developers, a build pipeline, or an obligation it believes the subscription discharges.
The honest summary
Two easy theses circulate and both are wrong.
"Overlays are snake oil and AudioEye is the worst of it" does not survive the evidence. The regulator fined a competitor, not AudioEye. AudioEye publicly repudiates overlay widgets. It ships a genuine developer SDK that tests staging and local environments and integrates with real test frameworks. It employs auditors and bought two audit businesses. It is the most transparent vendor in this category by a distance, because it has to be.
"AudioEye has moved past overlays" does not survive AudioEye’s own July 2026 documentation, which states that expert fixes deploy through the same JavaScript framework. That is an overlay by the Fact Sheet’s definition, however skilled the hands that wrote the patch. And 58 percent of revenue with essentially all the growth comes from a channel where the product is a pasted script at $328 a year, a price that cannot fund human audit work.
The version that survives is the one the filings show: two businesses, one brand, and a marketing layer that does not distinguish them. Read the quarterly report and you will know which one you are being sold.
Frequently Asked Questions
What is AudioEye?
A digital accessibility company founded in 2005, headquartered in Tucson, Arizona, and publicly traded on Nasdaq as AEYE. It combines a JavaScript product that remediates pages at runtime with human expert audits and a developer testing SDK.
What does AudioEye cost?
It no longer publishes prices. From its own filings, blended annual recurring revenue works out to about $328 per customer per year, or roughly $27 a month. Third-party sources quoting monthly tiers conflict by a factor of five to ten and should not be relied on.
Is AudioEye an overlay?
By the Overlay Fact Sheet’s definition, yes, and it is named there. AudioEye accepts that its JavaScript sits on top of site code but rejects the “overlay widget” category, which it says has “no business masquerading as an accessibility solution.”
Do the fixes stay if I cancel?
No. AudioEye’s own documentation states that expert-identified fixes deploy through the same JavaScript framework, so remediation is delivered client-side rather than committed to your source code.
Does AudioEye use real human experts?
Yes. It employs auditors and acquired two audit businesses. But with roughly 116 employees against 129,000 customers, that expert layer is concentrated in the enterprise segment, which is 42 percent of revenue and is not growing.
Was AudioEye fined by the FTC?
No. The January 2025 order and $1 million penalty over claims that an automated product could make sites compliant was against accessiBe. AudioEye was not named.
Was the detection study independent?
It was conducted by an outside agency and commissioned and paid for by AudioEye. The landing page discloses that and names the five tools tested. The press release that syndicated widely does neither.
Has anyone tested it independently?
We found no independent, non-commissioned controlled test of its remediation effectiveness published in 2025 or 2026. That absence is worth knowing before relying on any efficacy claim.
Can it test a staging site?
Yes, through a separate product. The Testing SDK and CLI can scan internal, staging and local pages, though the CLI does not support cookies or authenticated flows. That layer produces source-level findings and is not an overlay.
Which CMS platforms does it support?
WordPress, Shopify, Squarespace, Wix, Drupal, Duda, BigCommerce, Webflow, Joomla, HubSpot and others, installed by pasting a script into the footer. There is also a Finalsite integration.
Is there any case where an overlay is the right call?
The Overlay Fact Sheet’s own maintainer has written that one can offer a quick fix when problems are egregious and the organization is under outside pressure, as temporary relief alongside a funded plan for real remediation.
Should a nonprofit buy this?
If it has no developer, no build pipeline and no audit budget, roughly $27 a month buys detection and real runtime repair that would otherwise not exist. If it has developers or believes the subscription discharges an obligation, no.
