PDF remediation tools compared: Acrobat, CommonLook, axesPDF, Equidox and AccessLumens

We build one of the tools on this page, so read it with that in mind. What we have tried to do is write the comparison we wanted when we were choosing what to build against: what each tool is actually for, what it runs on, and what you are signing up to pay - including the things the others do better than us.

If you only need to know what is wrong, use PAC or veraPDF. Both are free, both are excellent, and neither appears in the table because neither repairs anything - they report. Run one whatever you choose to fix documents with. Adobe’s Auto-Tag API is absent on purpose too: it is a building block for developers rather than something you buy instead of a tool, and it bills 10 document transactions per page - a 20-page file costs 200 against a 500-a-month free tier, per its published usage limits.

PDF remediation tools compared by capability and by how each one is priced. Each capability cell reads Yes, Partial, or No.
Capability	AccessLumens	Adobe Acrobat Pro	CommonLook	axesPDF	Equidox
Runs in a browser, nothing to install	Yes	No	No	No	Yes
Works on macOS	Yes	Yes	No	No	Yes
Repairs the document automatically	Yes	Yes	Partial	Partial	Partial
Manual editing of reading order and tables	No	Yes	Yes	Yes	Yes
Priced per document, not per seat or per page	Yes	No	No	No	No
You only pay when the score measurably improves	Yes	No	No	No	No
What it costs	Low - $5 per document, no commitment	Moderate - per-seat subscription	Enterprise - quoted annually	Moderate - per licence	Enterprise - quoted annually

Yes
Partial
No

Things we do that most PDF tools do not

These are not in the table above, because scoring the other tools on them would mean guessing. Vendor documentation says “WCAG compliant” without saying what is checked automatically and what is left to the person operating it. So take this as a description of what we do, not a claim about what they do not.

Colour contrast, measured on the page as it is actually painted

We render each page and sample the pixels behind the text, rather than reasoning about fill colours and hoping they are what a reader sees. Contrast is the WCAG failure most PDF checkers hand back to you as a manual check, and Adobe documents that Acrobat does not determine whether your content meets it.

Descriptions for your images, drafted for you

A screen reader skips an image with no description. We read yours and draft a sentence for each one, shown beside the image so you can correct it before it goes in. Drafted, not decided - the words are yours, we just remove the blank page you would otherwise be starting from.

You see the measured score before you pay anything

The check and the repair both run free. What you are shown afterwards is the real re-measured score of the repaired file, not a projection of what a repair might achieve. If the number did not move, we say so, and there is nothing to buy.

The repaired file is checked again from scratch

Not a diff of what we changed - the same full check your original went through, so the before and the after are the same measurement of the same thing. It is the only way a before-and-after number means anything.

If the repair damages the document, we throw it away

An automatic repair can improve every technical measure while wrecking the page: fonts re-subset, text that no longer copies, a form that stops working. We compare the repaired file against your original page by page, and if it does not hold up we keep your original and charge nothing. We would rather return nothing than return something broken.

Your file is never written to disk

Uploads are held in memory for thirty minutes and then dropped. Nothing is stored, and there is no copy of your document to leak later.

On contrast specifically, Adobe is refreshingly direct about the limits of its own checker: Acrobat reports colour contrast as needs manual check and its documentation states that it does not identify whether your content meets the contrast requirement. That is not a criticism - contrast in a PDF is genuinely hard to measure, which is why we render each page and sample the pixels behind the text rather than trying to reason about fill colours. See Adobe’s own guidance.

The tools, one at a time

Adobe Acrobat Pro

Acrobat is the default, and for a lot of teams it is the right answer simply because it is already installed and already approved. It ships an accessibility checker, a full set of manual tagging tools, a reading-order editor, and the same Auto-Tag engine Adobe exposes through its PDF Services API - so it will make a first pass at structuring a document automatically and then hand you the tools to correct it.

The catch is not capability, it is throughput. Acrobat is built around one person working on one document, and the manual half of that work is slow whether or not the automatic pass went well. It is priced as a per-seat subscription, so cost scales with how many people are remediating rather than how many documents need it - which is comfortable for a team of two and painful for a backlog of two thousand.

If you have a handful of documents a month and someone with the patience to learn the reading-order editor, you may not need anything else.

CommonLook (Allyant)

CommonLook is the enterprise end of the market. It is a Windows desktop application built around guided remediation: it walks an operator through a document against a chosen standard - Section 508, PDF/UA, WCAG - and verifies the result at the end. That verification step is the reason it shows up in government and regulated-industry procurement, where being able to demonstrate conformance matters as much as achieving it.

It is the most thorough option here and also the most demanding. It expects a trained operator, it is Windows-only, and pricing is quoted on request rather than published. That is a statement of fact rather than a complaint - vendors who sell into procurement generally do not put a number on a web page - but it does mean you cannot evaluate the cost without starting a sales conversation.

Worth your time if you are remediating at volume with staff whose job this is, and if you need to produce evidence of conformance rather than just a better document.

axesPDF

axesPDF is aimed squarely at professional remediators, and it has a reputation for handling the structurally awkward cases - nested tables, complex lists, documents where the tag tree has to be reasoned about rather than generated. If you already know what a correct structure tree looks like, it gets you there with less fighting than the general-purpose tools.

It is a Windows desktop application sold per licence. Like CommonLook, it assumes expertise: it is a precision instrument rather than an automatic one, and the value is in the control it gives someone who knows what they are doing.

The wrong tool for a backlog you want cleared without hiring, and the right one for a specialist who needs to fix documents properly.

Equidox

Equidox is the closest thing here to our own shape: browser-delivered, so there is nothing to install and it works the same on a Mac as on a PC. Its documentation describes AI-driven detection that finds zones for text, headings, images, links and lists, table detection that identifies rows, columns and merged cells, and an integrated OCR engine for scanned documents.

Where it differs is the workflow. Equidox is explicitly human-in-the-loop: the detection produces zones, and a person reviews and adjusts them before exporting the tagged file. The company is direct about this, framing automation as the way to spend human effort on the judgement calls - alt text, reading order - rather than the repetitive parts. It also offers a separate higher-volume automated service. That mixed model is why it sits at Partial rather than Yes on automatic repair; the everyday product expects a reviewer.

Pricing is quoted rather than published. If you want a browser tool with a real manual editor and you have someone to operate it, this is the direct alternative to what we do.

AccessLumens

We are the narrow one. There is no manual editor: you upload a document, we repair what can be repaired automatically, and we re-check the result. That makes us a poor fit for a remediation specialist who wants control over every tag, and a good fit for someone with a pile of documents, no specialist, and a deadline.

The part that is genuinely different is what happens at the end. The repaired file is scanned again from scratch - the same check the original went through, so the before and after measure the same thing - and it is only kept if the score actually went up. If a document did not improve, there is no file and nothing to pay for. You see the real number before you decide, not a projection.

Priced per document repaired rather than per seat or per page, which means the cost tracks the backlog rather than the headcount. It also means a document we could not improve costs you nothing, which is the part of the model we would most like other people to copy.

How to choose

The honest split is not by quality, it is by who is doing the work. If you have a trained remediator, buy them a precision tool and let them use it - CommonLook or axesPDF exist because that work is real and skilled, and no automatic pass replaces someone who knows what a correct reading order looks like.

If you have documents and no specialist, the question becomes how much you can get fixed without hiring one. That is the gap we built for, and it is worth being clear about what falls in it: automatic repair clears structural barriers - tags, language, titles, reading order, embedded fonts - and it does not write your alt text or decide what a complex table means. Anyone telling you a tool makes a document fully compliant with no human involved is overselling, ourselves included.

Whichever way you go, check the result with something that did not produce it. PAC and veraPDF are free, and a tool grading its own homework is worth exactly as much as that sounds.

Try the one we build

Free to run and free to see the measured result. You pay only for documents whose score we have already measured as improved.

Repair a PDF

Every cell checked 2026-08-29 against each vendor’s own documentation. Only the shape of each price is given, never an amount we cannot source from the vendor - CommonLook and Equidox quote on request rather than publishing a rate. Products change; confirm with the vendor before you decide. Something here wrong or out of date? Tell us and we will fix it.