What the three things sold under this name actually are, how providers count what they charge for, and why two honest quotes for one site can land ten times apart. Most of this page applies whoever you hire, and the part about Harbor is the shortest section on it.
Three different products are sold under the phrase “accessibility audit”, and a first-time buyer usually assumes they are one thing. They are priced separately by most providers, they take different amounts of time, and only one of them is the document a procurement team is asking you for.
Minutes, often free
A rules engine crawls your pages and reports what a machine can decide on its own: a missing alt attribute, an unlabelled form field, contrast below the threshold. It is fast, repeatable, and genuinely useful as a first pass. What it returns is a list of detected issues, not a verdict on the standard.
Days of a person’s time
Someone operates your site the way a disabled user would, with a keyboard, a screen reader, and zoom, and works through the 55 Level A and AA success criteria of WCAG 2.2. Each failure is reproduced by hand before it is written down, and the finding says what to change rather than which rule fired.
The document procurement asks for
A VPAT is the blank template. Filled in, it is an ACR: a per-criterion conformance record attributed to whoever signed it. Procurement teams in government, education, and the enterprise ask for this by name, and an issue list will not substitute for it however thorough the issue list is.
The distinction matters most when you are reading a quote. A provider who quotes you a scan is not answering the same question as a provider who quotes you an audit, and neither of them has necessarily quoted you the report. All three are legitimate things to sell. They are simply not interchangeable, and the cheapest quote in your inbox is often cheapest because it is the smallest of the three.
Which of the three you need is usually decided by somebody else. If a customer, a grant, or an RFP has asked you for a VPAT, an issue list will not close that request. It is also a document most software companies do not have sitting ready: of 107 recognizable SaaS vendors we checked on one day, 30 published a conformance report a buyer could read without asking anyone for it, which is 28%. The full finding is published.
A scan also belongs inside a real audit rather than opposite it. It clears the mechanical failures quickly and tells a person where to start looking. The narrow claim in the next section is about what a scan can decide on its own, which turns out to be the smaller part of the standard.
WCAG 2.2 has 55 Level A and AA success criteria. That is the list a conformance report answers, one row at a time.
Harbor’s pre-registered study methodology classifies roughly 30 of those 55 as not machine-testable to an evidence standard, including the whole 1.2.x series on media alternatives, the semantics of focus order, and the quality of error suggestions. The rest split two ways: some a tool can decide outright, and some it can only flag for a person to confirm. Each criterion sits in exactly one of those three groups, so the three account for all 55 with nothing left over.
Where exactly the line falls is a matter of how strict an evidence standard you hold a tool to, and reasonable people draw it a few criteria either side of 30. The shape holds wherever it is drawn: the majority of the standard needs a person, and the reason is easier to see in examples than in a proportion. Each of these passes an automated check and fails a human one.
None of that makes scanning worthless, and a provider who runs one is not cutting a corner. It sets the ceiling on what a scan alone can support. A tool narrows the search, and a person has to finish it before anyone signs a document that says a criterion is supported.
This is the most useful thing on the page, and it is not complicated. Providers charge for a count, and they do not all count the same thing. Three models are in common use.
Per page. The fee is your page count times a rate. Published rates run from $100 to $350 per page, and definitions differ: some providers price “primary pages” and test a representative sample rather than the whole site, which is a materially different number from the same words.
Per template. The fee follows the number of distinct layouts. Three hundred products rendered by one product template count once, because testing the three hundredth teaches you what the second one already did.
Tiered. A fixed fee for a bundle, with caps. Read the caps closely. One published tier table caps pages and templates together at each level: 5 pages with 2 templates, then 10 with 4, 30 with 12, 50 with 20, and 100 with 40. You land in the tier set by whichever cap you hit first, which is often not the one you were shopping for.
Put one site through all three, reading the per-page model both ways it gets used. A store with 300 products, 20 category pages, and one each of home, cart, checkout, and account has 324 pages, and those 324 pages render through 6 layouts.
| Pricing Model | What the Fee Counts | The Store Above |
|---|---|---|
| Per page, whole site | All 324 pages, at $100 to $350 each | $32,400 to $113,400 |
| Per page, sampled | 12 representative pages, at the same rate | $1,200 to $4,200 |
| Per template | The 6 layouts, with the catalogue excluded | Six templates, at any catalogue size |
| Tiered, with caps | Whichever cap binds first, pages or templates | Set by the page cap, not by the 6 layouts |
Rates are the $100 to $350 per page figures published by providers on the comparison page, read from their own pages on August 4, 2026. The store is hypothetical and the arithmetic is shown so you can check it: 324 pages times $100 is $32,400, and times $350 is $113,400.
Sensible providers do not bill 300 near-identical product pages one at a time, so the first row is a boundary rather than a quote you would receive. It is in the table because it is what a per-page rate literally says, and because the gap between the first row and the second is the thing to pin down before you compare any two prices. The same store is 324 pages, 12 pages, or 6 templates depending only on who is counting, and each of those is a defensible basis for a fee.
So a low quote is usually a small scope rather than a shortcut, and a high one is usually a larger count rather than a markup. You cannot tell which from the number, and the number is the only part most providers put on a public page. Ask all of them the same two questions: what counts as a page here, and how many are in scope. Then compare.
One more variable sits on top of the count. A per-template fee with no page cap and a tiered fee that counts pages behave in opposite directions as your catalogue grows, so the cheapest provider for a five-page brochure site is frequently not the cheapest for a store, and the ranking flips somewhere in between.
The published prices below are the ones to run your own page and template counts against, including the shapes where a cheaper provider wins.
Each figure below was read from the provider’s own published page on August 4, 2026. These are list prices rather than quotes, and a published range is a starting point rather than a promise, so treat the table as a way to calibrate the number you have been sent rather than as a menu.
| Provider | Published Price | How Page Count Affects It | Conformance Document |
|---|---|---|---|
| Accessible.orgaccessible.org/pricing | $1,250 to $2,750 | Priced per page | Sold separately, from $350 |
| Skynet Technologiesskynettechnologies.com | $500 to $12,500 current | Pages AND templates capped | Included in the VPAT tiers |
| DigitalA11Ydigitala11y.com/pricing | $100 to $350 per page | Priced per page | Sold separately |
| TestProstestpros.com | $3,000 to $10,000 | Not published | Included |
| Web Audit Solutionswebauditsolutions.com/vpat-services | VPAT from $500 | Tiered by page count | Sold separately |
The detail behind the price column is where the comparisons break down. Accessible.org publishes $100 to $250 per primary page and states that most audits cost $1,250 to $2,750, and prices the conformance report separately by edition, from $350 for the WCAG edition. DigitalA11Y publishes $100 to $350 per page and quotes the VPAT separately at $1,000 to $3,000. Skynet Technologies publishes tiers from $500 to $12,500 with the dual caps described above. TestPros publishes $3,000 to $10,000 as a range with no scope breakdown attached. Web Audit Solutions prices audit tiers by page type and prices the VPAT by page count, from $500.
Say the obvious thing about that list out loud: the conformance document is frequently billed on top of the audit, at anywhere from a few hundred dollars to several thousand among the providers who publish a figure for it at all. An audit price and a report price are two numbers. If you compare a quote that bundles them against a quote that does not, the bundled one looks expensive for a reason that has nothing to do with the work either of them would do.
The table is also not the whole market. It only includes providers who publish enough to be compared, so a provider absent from it may simply not post prices. Harbor is at the bottom of the table because this section is about the category, and because we are the most expensive entry price on it.
Figures come only from each provider’s own pages, never from a review site or a pricing aggregator, and are re-checked quarterly. Follow the source link in any row to check it yourself. Prices may have moved since August 4, 2026, and if you believe something here is inaccurate, tell us and we will correct it.
Send these to everyone you are considering, in the same words, and compare the answers rather than the prices. Two of them are questions we would have to answer carefully ourselves, which is why they are here.
“What counts as a page, and how many pages or templates are in scope?”
A good answer: A number and a definition, before the fee is agreed, ideally with the sampled URLs or the template names listed. A quote with no count attached cannot be compared to any other quote, and this is the single question that explains most of the spread between two prices.
“Is the conformance report included in this fee, or billed separately?”
A good answer: A plain answer with the figure and the edition, since Section 508, WCAG, and international editions are priced differently by the providers who publish them. If it is separate, get that number now rather than after the audit lands.
“How much of the testing is manual, and against which version and level?”
A good answer: WCAG 2.2 Level AA is the common target, and the answer describes the manual work: keyboard, screen reader, zoom and reflow, and which assistive technology on which browser. An answer about which tools they license is an answer about software, not about method.
“Who performs the evaluation, what are their credentials, and what happens if they leave mid-engagement?”
A good answer: A name or at least a role and a credential. Answering at the company level is normal and not a red flag. Ask the second half either way: a firm with several testers absorbs an absence, and a one-person practice, ours included, is a single point of failure you should price in.
“How long have you been doing this, and how many of these have you delivered?”
A good answer: A straight number. A new practice and an established one can both do good work, and the honest answer from a new one is a small figure plus samples you can judge yourself. Ask for the samples in both cases, because the work is better evidence than the count.
“Can I read a full example of what I will receive, before I sign?”
A good answer: A complete sample, redacted if it belongs to a client. Read how the findings are organised. Grouped by root cause with the code to change means your developers can work from it. One row per instance, sorted by success criterion, is closer to an export than a plan, and it will cost your team the difference in triage.
“Is a re-test after we fix things included, how many rounds, and is the report re-issued?”
A good answer: A number of rounds and a clear answer on the document. Practice varies widely: some providers include a round, and at least one publishes validation as a separate item priced at half the original audit fee. Both are legitimate. You just want to know before the invoice.
“When will it be delivered, and what happens if that date moves?”
A good answer: A date, and what follows if it slips. Most published pages give an estimate rather than a commitment, which is normal in this market. The reason to ask is to find out whether the date is a plan or a promise before your own deadline starts depending on it.
A provider who cannot answer one of these on the spot is not disqualified. Several of these are commonly settled in a contract or on a call rather than published on a website, so treat a gap as something to raise rather than as a verdict.
The delivery date a provider quotes you is the date of the report. It is not the date your site conforms, and the distance between those two is mostly on your side of the fence. Expect the engagement to run in this order.
What that means practically is that an audit needs your engineering time, and buying one without it produces a document rather than an outcome. Budget for a developer who can act on findings, and for someone who can say what a screen is supposed to do, because a fair number of judgement calls turn on intent that only your team knows.
If that time is not available this quarter, buying the audit anyway is still a reasonable decision. You get a prioritised plan and a conformance document that describes where you actually are. Just go in knowing the report will say that, so nobody in the room is surprised by an honest one.
Two things worth agreeing in writing before you start: how many re-test rounds you get, and whether the report is re-issued after each one. Those two lines decide whether the engagement ends with a document that matches your fixed site or one that matches the site you had at kickoff.
Briefly, because the rest of this page is the useful part. Harbor is a one-person practice. Audits are performed under the DHS Trusted Tester for Web certification, number TT‑2606‑09221, and your report carries that name and number. We price by template with no page cap, and the conformance document is inside the fee rather than beside it. A Quick Check covers up to 5 pages and returns a findings report, which is not a signed conformance document.
We do not publish a figure, which is why we are absent from the table above rather than sitting at the bottom of it. The fee follows your unique template count, and fifteen minutes on a call is enough to establish it and fix it in writing before you commit to anything. By the arithmetic in this guide that tends to make us dearer than a sampled per-page quote and cheaper than a per-page fee once your page count grows, and we are a single practitioner rather than a firm with a bench. Which of those matters most depends on what you are buying and on who is going to read it.
The tiers and what sits in each are on the engagement page. The row-by-row version of the table above, with sources and the rows we lose, is the named comparison. We no longer publish our own figure, so we are absent from the market prices above rather than quietly favoured by them: ours is agreed on a fifteen-minute call and fixed in writing before you commit.
The eight questions above are the fastest way to find out whether a price is fair. Send us the scope if you want a second opinion on what to ask, whether or not we are on your list.
Competitor figures on this page were read from each provider’s own published pages on August 4, 2026 and are re-checked quarterly. They are reproduced for comparison and may have changed since. Harbor research cited here is published in full at the ACR availability finding.