Guide
What PDF remediation actually costs
Vendor rates, what makes one document ten times another, what per-page pricing hides, and how to tell whether outsourcing or software is cheaper for you.
About 11 minutes · quick
What the market charges
Outsourced remediation is sold per page. Quoted rates generally sit somewhere between USD 3 and USD 15 per page for ordinary documents, with a separate and higher band for forms, dense tabular data, mathematics, and non-Latin scripts. Volume discounts are normal; rush surcharges are normal; minimum engagement sizes are normal.
Software is sold per seat per month. Taggart is USD 29 or 59 a month depending on the plan, with checking free and unlimited on every plan including none. axesPDF is around USD 650 a year. CommonLook and PDFix sell per seat and per volume respectively; several tools are free and check without fixing.
Neither number tells you what a document costs you. This does.
What actually drives the cost
The variance between documents is enormous and almost entirely predictable from four things.
Does a structure tree exist, and is it any good? A PDF exported from Word with real heading styles arrives with a usable tree. The same document with headings that are just bold 16pt text arrives with a tree full of P elements and no outline, and somebody has to decide which of those paragraphs are headings. That single difference is roughly a factor of three.
How complex is the layout? Single-column body text is the cheap case: the reading order is already right and there is little to correct. Multi-column pages, sidebars, pull quotes, floated figures with captions — each is a decision about where something belongs in the reading order, and reading order is the most expensive step in the process.
How many tables and figures? Every figure needs alt text somebody writes. Every table needs its headers identified and associated, and a table with merged cells or stacked header rows needs explicit Headers references, which in Acrobat is one dialog per cell. A page with a single complex table can cost more than ten pages of prose.
Is it a form? Forms are the most expensive documents in the category. Every control needs a tooltip somebody writes, a Form element in the structure tree, and a position in the reading order — and in Acrobat that is a five-step process per field, thirty times on a thirty-field form. A form is not a document with fields on it; it is a different job.
Two things sit outside the scale entirely. A scanned document has no text to tag and needs OCR before any of this begins — and OCR quality decides everything downstream. A document whose source you still have should usually not be remediated at all: fix the source, re-export, and the cost approaches zero for every future version.
What per-page pricing hides
A per-page rate is an average over a distribution the buyer cannot see. Three consequences are worth knowing before you sign.
Your simple documents subsidise your complex ones. If your backlog is mostly plain reports, you are paying the blended rate for work that is cheap.
“Compliant” may mean “passes a checker”. Automated checkers verify that alt text exists, not that it says anything; that a reading order exists, not that it is correct. A vendor optimising for a green checker result can deliver a document that validates and reads as nonsense. This is exactly what to look for when reviewing delivered work.
The judgement calls are invisible. Somebody decided which cells are headers and what the chart conveys. If those decisions are not written down, you have no way to audit them, and neither does a regulator. That record is what a remediation report is for, and it is the single most useful thing to insist on in a contract.
Working out your own number
Do this before getting quotes, not after.
1. Count the backlog, by type. Not “3,000 PDFs” — how many are forms, how many are scans, how many are the same monthly report with different numbers in it. Triaging a backlog is a day of work that routinely removes a third of it.
2. Find the duplicates and the templates. A report produced monthly from the same template is one fix at the source and then nothing, forever. This is usually the largest single saving available and it is almost always missed.
3. Check a representative sample. Take twenty files across the types, run a free check on each, and look at what comes back. A file failing on metadata and language alone is ten minutes. A file with no structure tree is an afternoon. You now have a distribution rather than an average.
4. Time one of each type yourself. Remediate one plain document, one with tables, one form. Whatever it takes you, that is the honest input — and it is a far better estimate than any rate card, because it is your documents and your standard.
5. Then compare. Outsourcing costs (pages × rate). Software costs (licence) + (hours × your loaded rate). The second only wins if you have somebody who can make the judgement calls — and if you do not, note that outsourcing is buying that person, not avoiding them.
Where software changes the arithmetic and where it does not
The mechanical majority is genuinely mechanical: metadata, language, artifact nesting, role maps, heading level repair, table scope where headers are already marked, placing form controls in the tree. A tool should do all of that unattended and in seconds, and paying a per-page rate for it is paying for keystrokes.
The rest does not compress. Somebody still decides which paragraph is a heading, what the chart conveys, where the sidebar belongs, which cells are headers. Software can put each decision in front of a person with the evidence attached and record what they chose — which makes the decision fast and auditable — but it cannot make it.
So the realistic saving is not “remediation becomes free”. It is that the mechanical part goes to zero and the judgement part becomes the whole job. On a typical backlog that is a large fraction of the time, and it is concentrated in exactly the documents that are cheapest to begin with.
What to put in a contract
Whether you buy software or a service, four clauses make the result verifiable:
- A named standard. WCAG 2.1 Level AA, and validates against PDF/UA-1 (ISO 14289-1) as the acceptance test. Why both.
- A named validator, so “compliant” has an operational meaning and both sides can run it.
- A remediation report recording what was found, what was changed, and what a person decided and why.
- A statement of what was not done — the conditions that need human judgement and were left open, and why.
The fourth is the one nobody offers and the one that tells you most about who you are dealing with.
Check a PDF free → · Pricing → · Triage a backlog →
Frequently asked questions
What is a typical per-page rate?
Published and quoted rates for outsourced remediation cluster between roughly USD 3 and USD 15 a page, with complex documents — forms, dense tables, mathematics, right-to-left scripts — quoted separately and often several times higher. Treat any single number as an average across documents that genuinely differ by a factor of ten.
Why do vendors price per page at all?
Because it is the only unit a buyer can count before the work starts. It is a poor proxy: a page of body text is minutes and a page of nested tables is an hour, and the vendor absorbs that variance by averaging — which means simple documents subsidise complex ones.
Is software cheaper than outsourcing?
It depends almost entirely on volume and on who is doing the judgement work. Software removes the mechanical majority and leaves the decisions; if you have somebody who can make those decisions, software is dramatically cheaper past a few dozen documents. If you do not, you are buying that person's time either way.
What should a quote include?
A named standard, a named validator, a remediation report recording what a person decided and why, and a statement of what was not done. A quote that says "WCAG compliant" with no test and no report is unverifiable, and you will find out at audit.
Does a cheaper rate mean worse work?
Not reliably, but very low rates usually mean auto-tagging with light review, and auto-tagging fails hardest on exactly the documents that matter most — forms, tables, multi-column layouts, and anything not in English. Ask what the process is rather than what the rate is.
Related failure conditions
- 09-001 Tags are not in logical reading order The order of elements in the structure tree does not match the order a human would read the page. This is the single most common substantive defect in tagged PDFs and the one Acrobat's own tooling handles worst.
- 15-002 A column has a header cell, but that header cell is not tagged as a header The top row labels the columns but is tagged as ordinary data cells.
- 13-001 Graphics objects are not tagged with a Figure tag An image or vector drawing that carries meaning is not inside a Figure element. Either it is untagged, or it has been swept into a paragraph.
- 28-010 A widget annotation is not nested within a Form tag Every form control must sit inside a Form structure element.
Check your own file. Taggart's validator is free and unlimited — no page cap, no watermark, no account.
Last updated .