AccessLumens
AI-based automatic repair

Fix an inaccessible PDF

Upload your document and get a repaired one back. We rebuild the structure a screen reader needs, set the language and the title, and draft descriptions for your pictures with AI. Then we check the result — and you only pay if it actually improved.

Drag & drop PDFs here, or click to browse

Up to 5 PDF files, 10 MB and 300 pages each.

Free check • First complete report free • Files are never stored, discarded right after the check

Free to run and free to see the result. $5 per document for the repaired files - and nothing at all for a document whose score does not move. No sign-up. Your document is held in memory and never saved to disk.

+50

average score improvement

across nine real documents

8 of 9

documents improved

the ninth was free

0

documents damaged

every repair is re-checked first

What we fix

The structure a screen reader follows

Most inaccessible PDFs have none at all, which is why a screen reader reads them as a jumble or not at all. This is the single biggest thing we fix.

The language of the text

Without it a screen reader guesses, and English read with a Spanish voice is close to unusable. We work out the language from the text rather than assuming.

A real title, not the file name

So the document announces itself as "2026 Budget" instead of "Microsoft Word - final_v3_FINAL.docx".

Descriptions for your pictures, drafted with AI

A screen reader skips past a picture with no description. We read yours and draft a sentence for each, shown next to the image for you to check and place.

The order the Tab key moves in

So someone navigating by keyboard moves through the page the way it reads, not the order things happened to be added.

Fonts that were never embedded

So the document reads the same on a machine that does not have the fonts it was written with.

What we will not do

No tool makes a PDF accessible in one click. Here is what we leave to you.

  • Reading order. We show you where it is wrong. Rearranging meaning is your call.
  • Image descriptions. AI drafts them, you place them. A wrong one is worse than none.
  • Bookmarks. We will not invent a contents list a document never had.
  • Colour contrast. We name what fails and what would pass. We never repaint your document.
  • Tables and lists. What a column means depends on what it says, so guessing reads as nonsense.

All of it is in your report, with where it is and what to do.

How we prove it worked

We check the document again from scratch

The repaired file goes through exactly the same check your original did, so the before and after are the same measurement of the same thing.

We compare how it looks, page by page

An automatic repair can improve every technical measure while visibly wrecking the page. If the repaired document does not look like yours, we throw it away and keep your original.

You see the real number before you pay

Not an estimate. If the score did not move, we say so and there is nothing to pay for.

How this compares to Acrobat, CommonLook, axesPDF and Equidox

A side-by-side of what each tool does, what it runs on and what you are signing up to pay - including the things they do better than us, and why the two free checkers worth running are not in the table.

Compare the tools

Common questions

What is PDF remediation?
Repairing a PDF so assistive technology can read it: giving it the underlying structure a screen reader follows, setting the document language and title, describing images, and making the keyboard move through it in a sensible order. Some of that a machine can do reliably; some needs a person. We do the first part and tell you exactly what is left, including the image descriptions, which today you write yourself.
Will it change how my document looks?
No. We check: every repaired document is rendered page by page and compared with your original, and if it does not look the same we throw our version away and keep yours. That check exists because an automatic repair can improve every technical measure while visibly damaging the page.
What if it doesn't work?
Then you pay nothing. We run the repair for free, re-check the document afterwards, and only ask for money against an improvement we have already measured. On our own test set one document in nine could not be improved at all, and that one would have been free.
Is my document stored?
No. It is held in memory while we work and never written to disk. After about thirty minutes it is gone, and you would need to upload it again.
Does this make my PDF fully compliant?
Usually not on its own, and any tool promising otherwise is overselling. Automatic repair clears the structural barriers, which is most of the distance for most documents. What remains needs a person, and we list exactly what and where.

Want to see what is wrong first?

The checker gives you a full audit against WCAG 2.2 and PDF/UA, with every issue and where it is. You can repair it from there too.