Nobody can tell you what fixing your site will cost, and anyone who names a figure before looking at it is reading you their own rate card. No price index exists for this work. No published average, no benchmark, nothing a buyer can hold a quote up against. So this page does the next best thing. It hands you the four things that decide the number, which is enough to work out your own before you ask a single vendor for a price.
Remediation is the industry's word for the fixing, as opposed to the finding. The two are separate purchases and they get muddled constantly, usually by whoever is selling both. What an audit costs covers the finding, where prices are published and roughly comparable. This page covers the fixing, where they are neither.
The Only Public Numbers, and Who They Actually Describe
One set of real figures sits in public, and it got there through a federal rulemaking record rather than through the market. American local government was asked what this work was costing, and the answers went in on the record.
- One-time quotes for PDF and website remediation ran from $10,000 to $20,000.
- Annual remediation estimates reached as high as $70,000.
- A city of about 275,000 people put its own effort at roughly 1,300 staff hours, covering training, auditing, remediation oversight and enforcement support.
County government added the more useful half. Federal cost estimates were accurate or on the high side, they still came in above what counties had budgeted, and the hardest single part of the job was remediating PDFs. Not the website. The documents hanging off it.
Now the part that decides how much use those numbers are to you. They describe US municipalities with large public document estates, a statutory deadline and salaried staff to redeploy. They do not transfer to a private business or a nonprofit, and we are not going to pretend they do. A ten-page brochure site and a city of 275,000 people share a standard and share nothing else. What the figures are genuinely good for is the shape they show when somebody counts honestly. A heavy documents bill, a long tail of oversight, and an annual number that never quite reaches zero. The full context sits on our page for the HHS Section 504 web rule.
Why There Is No Price List, and Why There Cannot Be One
This looks like an immature market and it is not. It is structural, and the two reasons underneath it will still be true in ten years.
The first is that there is no unit to price. Conformance in the standard is defined for a web page and never for a website, so "getting the site fixed" names nothing the standard can measure. What you are actually buying is some number of pages, each of which conforms or does not on its own, and the count depends on decisions nobody has made yet. Our guide to what W3C's evaluation methodology requires has the long version, and it is worth reading before you sign anything that promises a compliant site.
The second is that two sites with identical page counts can be a week apart or a year apart, and the thing that separates them is invisible from the outside. One has four templates and a design system. The other has eleven years of pages built by six agencies. A quote written without opening either is a guess with a number on it.
The Four Things That Decide Your Number
Four questions, and you can answer all four yourself. How many templates you really have. Whether the failures live in the template or in the content. Whether your documents are in scope. And whether the people doing the work have done it before. The rest of this section takes them in that order, because that is roughly their order of impact.
How Many Templates, Not How Many Pages
Page count is the number everyone reaches for and it is close to meaningless. Four hundred blog posts sharing one template is one piece of work. A home page, a category page, a product page and a checkout is four, and each one needs its own pass.
The reason this matters so much is that one bad decision reports as many failures. WebAIM's 2026 sweep of a million home pages found low-contrast text on 83.9% of them, averaging 34 separate instances on a single page. That is usually one color choice, counted 34 times. Six error categories accounted for 96% of everything detected across the whole million. So a thousand-row scanner export is not a thousand problems, and pricing the fix off the row count is how people talk themselves into a number ten times too big.
One thing to add to your template count before you finish it. The standard treats each layout a responsive page presents at a different screen size as a variation that has to conform on its own. A template whose mobile arrangement is genuinely different, rather than the same arrangement narrowed, is closer to two templates than one.
Whether the Failures Sit in the Template or in the Content
This is the split that decides whether you have a project or a habit to change, and it is the single most useful thing you can learn in week one.
Template failures are fixed once and inherited everywhere. Color tokens that miss the ratio, a focus style somebody removed, a heading level skipped in the layout, form markup in a shared component. Your developer fixes the component and every page that uses it comes good in the same deploy. This half is bounded, it is estimable, and it is the half a quote can honestly price.
Content failures are fixed one item at a time, and they come back with the next thing you publish. Alt text that describes the wrong thing. Link text reading "read more" forty times. A heading typed as bold text because bold looked right. A video with no captions. Nobody fixes this half once, because the people creating it are still creating it, which is why training your team and keeping the site accessible do more for that number than developer hours ever will.
So when a quote arrives, the first thing to check is whether it separates the two. A single figure covering both is either pricing the content work as though it ends, which it does not, or quietly excluding it, which you find out about later.
Whether Your Documents Are in Scope
A PDF you link to from your site is a web page for conformance purposes. The standard defines a web page as a non-embedded resource obtained from a single address over HTTP, and that is exactly what a PDF is. Which means it carries the same rules your HTML does, and nobody enjoys hearing it.
Here is why documents blow budgets. There is no template to inherit from. Two hundred PDFs is two hundred separate jobs, each needing its tags, its reading order, its table headers, its alt text and its title, and none of that work makes the next document easier. The PDF world has its own conformance specification and its own published failure list, the Matterhorn Protocol, which runs to 31 checkpoints and 136 ways to fail. This is a discipline with its own tooling rather than a long afternoon in Acrobat, and our guide to PDF accessibility covers what the work involves while how to tag a PDF walks the process.
This is why counties named documents as the hardest part, and it points at the cheapest decision available to you. Go and count your PDFs, then work out how many of them still need to be documents at all. A price list, a menu, a form and a policy are usually better as web pages anyway. A document you turn into a page stops being a remediation line item and starts inheriting your template fixes for free.
Whether the People Doing the Work Have Done It Before
A practiced developer fixes a labelling problem in minutes. A developer meeting the subject for the first time spends an afternoon on it, and there is a decent chance the fix introduces something new.
That last part is measurable. In the same million-page sweep, pages using ARIA averaged 59.1 detected errors against 42 for pages without it. That does not prove ARIA causes failures, because complicated applications use more of everything. What it does kill is the instinct that adding attributes is how you clean up a report, and that instinct is exactly what an inexperienced team brings to a findings list. Our explainer on ARIA is the antidote and it is free.
The practical read is that your first remediation costs more than your second, whoever does it, and that most of the difference is knowledge rather than hours. Which is the whole reason our 432 fix guides are public and ungated. A report that links each finding to the fix is a report that prices this factor down, and we would rather your team billed you for the fixing than for the learning.
How to Put a Number on Your Own Job in an Afternoon
Six steps, in order, and none of them needs a vendor. The output is a scope you can hand to three firms and get comparable prices back, which is the thing you actually want.
- Count your distinct templates. Open your site and list the layouts, not the URLs. It is almost always fewer than you expect.
- Add one for every template whose phone layout is a different arrangement, rather than the same one narrowed. Those have to conform on their own.
- Write out your key processes, step by step. Every page in a process has to conform for the process to conform, and the standard's own worked example is a store checkout. A four-step checkout is four pieces of work, not one, and it is the thing quotes underestimate most often.
- Count the documents you actually need people to read, then mark the ones that could be web pages instead. The second number is your saving.
- Run a free scan across ten pages and read what repeats. A failure appearing on all ten is a template failure and it is the cheap half. A failure appearing once is content, and content is the half that keeps coming back.
- Do the keyboard pass yourself. Put the mouse away and try to buy something or finish your main form. Ten minutes, no tools, and it finds the blockers no scan reports.
Then ask for a price per template and per document rather than for the site. A vendor who can quote that way has understood the job. A vendor who insists on one number for everything is either carrying a lot of risk premium or has not looked yet, and both of those come out of your budget.
How Long It Takes, and the One Public Timeline
The honest answer on time has the same shape as the honest answer on cost, so here is the fullest public record of one running end to end that we have found.
Norway's accessibility regulator inspected a regional health portal used by about 425,000 people. Testing ran from late April to late June 2025 across 19 pages, and the report landed that August recording 119 individual errors, in 12 of the 14 requirements tested. The portal was ordered to fix them. At the correction deadline in early December, 64 errors were still open. The regulator set a fresh deadline later that month and attached a fine of 50,000 kroner a day. At that deadline six errors remained, the fine started, and four days later the work was done and the case was closed. The fine ran for two days.
Read that for its shape rather than its duration. The count did not fall steadily. It sat, and then it moved, and it moved when the deadline had money attached to it. That is one public body under a regulator, so it is not a forecast for your site. The part that does transfer is not about speed anyway. It is that the end of a findings list is where remediation stalls, which means the tail needs a date on it from the start rather than an intention.
There is a second piece of evidence on timing and it comes from the government that wrote a deadline. HHS pushed both of its Section 504 web compliance dates back by a full year. Its stated reason was that a significant number of the organizations covered, especially local governments and smaller recipients, were not going to make them. When the regulator moves its own date, planning as though your first estimate holds is optimistic.
What you can plan against is the split from earlier. Template work is bounded and usually measured in weeks, because it is a finite list of components. Document and content work is measured in decisions about what you publish and who is allowed to publish it. That one never finishes, and that is a fact about the work rather than a sales position.
Four Questions That Make a Remediation Quote Readable
You will not be able to compare two remediation quotes on price, because they are almost never quoting the same job. You can compare them on these.
- Is this priced per template or per page? Per page on a templated site means somebody is either padding or has not looked.
- Are the documents in it, and how many? If the number is missing, the documents are out, whatever the covering email says.
- Who verifies the fixes, and is it the same people who made them? Marking your own homework is not verification, and a firm that audits, fixes and then signs off its own work has no independent step in it anywhere.
- Who owns the code, and what happens if we leave? A fix that lives inside a vendor's script disappears when you stop paying, which is the overlay problem wearing different clothes. Our record on overlays is the long version.
Where we stand in this
We sell the audit, at a published price, and the fixing is done by your own team or by a supplier you pick, working from the report and the free fix guides. We do not sell repair work, at any price, in any form, before or after the report. So the findings list can never double as a sales pitch for the fixes, the number on the pricing page can be fixed before we look, and when we come back and retest, we are checking somebody else's work rather than our own. That arrangement costs us real money every month, and it is the only reason any of those three sentences is true.