Drop your PDF into the zone below and get back a file built to the archiving standard that public tenders, courts and long-term records systems ask for by name. Fonts are embedded, the color profile and metadata the standard requires are written, and anything executable is removed. Then the file is audited, and if something still fails, this page tells you what and why instead of stamping the label on and wishing you luck. Nothing is uploaded.
Your files are read by JavaScript running on your own machine. There is no upload, no queue, and no copy on our server to delete later.
Drag a card to reorder it, or use the arrows under each card. The order here is the order in the output. Each card is a preview of the finished page, so what you see is what the PDF will contain.
Esc close ← → move R rotate
Starting
This is your processor doing the work, so the time it takes depends on your machine and not on how many other people are on the site.
The converter above is ready to use. The guide below is the part that decides whether the file you produce actually passes validation - what the standard really requires, why the fonts are the hard part, which of the three parts you should choose, and what to do when a document cannot be made to conform without giving something up.
Three minutes, nothing to install, and no account needed for your first run. This PDF/A converter executes inside the page you are reading, so every step below happens on your machine - which matters here more than on most tools, because the documents that get archived are the ones nobody wanted to upload.
Click the drop zone above or drag the files anywhere onto this page. Each document is parsed as it lands, so a password protected or damaged file reports the problem before you spend an operation on it. Encrypted files are refused outright. The standard forbids encryption, and a password on an archived document is a key somebody will eventually lose, which defeats the whole point of archiving it.
PDF/A-2b is the right default and what most archives and tenders now specify. Choose PDF/A-1b only when a system asks for it by name, because it is the strictest of the three and forbids transparency outright, which trips up any document carrying a modern logo. Choose PDF/A-3b when the file has to carry an attachment. That is what electronic invoicing formats use.
Preserve text keeps the document searchable, embeds substitute outlines for any standard font that was left out, and reports anything it could not fix. Rebuild as images rasterizes every page, which guarantees conformance because a document with no fonts cannot violate the rules about fonts, and costs you the text layer. Fill in a title too. The standard requires the metadata and the document information to agree, and both are written from that field.
Press Convert to PDF/A and the work happens on your processor. The result screen states plainly whether the file meets every requirement this converter can check, or names exactly what still fails and why it could not be fixed. Read it first. A file that claims conformance and fails at the archive door is worse than one that never claimed any.
PDF/A is a subset of PDF, standardized as ISO 19005, and it is not a compression setting or a checkbox you tick. It exists for one reason. A document opened in eighty years has to render identically to the day it was written, and every rule in the specification follows from that goal, which in turn follows from a single principle: everything the page depends on must be inside the file, and nothing may depend on the reader's environment.
In practice that is four requirements and a long list of prohibitions.
Why this tool audits before it converts. Almost every free converter writes the metadata declaration and calls the file PDF/A. That produces a document that fails validation at the archive's door while claiming to pass - and the failure is discovered by a records officer or a regulator rather than by the person who could have fixed it. This one checks first, fixes what a browser can fix, and refuses to claim conformance it did not achieve.
The three parts are three different base versions of PDF with three different sets of what is permitted, and the right one is the one your recipient asked for. A higher number is not better. It is only newer, and newer permits more. Three kinds of obligation drive almost every PDF/A request a normal business ever receives: a court or tribunal filing rule that names the format in its electronic submission guidance, so a brief filed as an ordinary PDF bounces at the portal before a judge ever sees it; a records retention schedule, meaning the rule that says how long a document must be kept and in what form, set by a national archive, a hospital, a university library or your own compliance officer, where the period routinely runs to seven, ten or thirty years; and an invoice archiving law, since several European countries require a business to keep the invoices it issues and receives in readable form for a decade, which is exactly the job part 3 was built for. If none of those apply to you, ordinary PDF is fine. Nobody asks for PDF/A casually, and the request almost always arrives with a part number attached, so use the part your recipient named rather than the newest one on the list.
| Part | Built on | Allows | Choose it when |
|---|---|---|---|
| PDF/A-1b | PDF 1.4, from 2005 | No transparency, no layers, no attachments, no object streams. | A system asks for it by name. Some court filing systems, some national archives and a lot of older procurement software still specify it and will reject anything else. |
| PDF/A-2b | PDF 1.7, from 2011 | Transparency, layers, JPEG 2000, and PDF/A files embedded inside it. | Almost always. It is the current default across European public procurement and most institutional archives, and it accepts documents part 1 would reject. |
| PDF/A-3b | PDF 1.7, from 2012 | Everything part 2 allows, plus arbitrary file attachments. | The document must carry a machine readable payload. This is what electronic invoicing formats use to put structured invoice data inside a human readable PDF. |
Every part has conformance levels, and the letter matters as much as the digit. This converter produces level B, which means "basic" and guarantees visual reproducibility - the file will look the same forever. Level A goes further. It requires a full logical structure tree: tagged headings, reading order, alternative text on every image, and language markers, so the document is accessible and its content is machine readable in the semantic sense.
Level A cannot be produced from a flat PDF by any tool, here or anywhere, because the structure it demands does not exist in the file. Tagging has to be added by the software that created the document, or by hand afterwards in a full editor. Any converter offering to turn an untagged PDF into PDF/A-1a is either adding a structure tree it invented or lying about the level. If your specification says level A, the fix belongs upstream in whatever produced the PDF. Not in a converter.
Every requirement in the standard is straightforward except one. Color profiles, metadata, identifiers and the removal of scripts are bookkeeping. Fonts are not. Font embedding is a genuine engineering problem, and it is where the overwhelming majority of PDF/A conversions fail, including plenty performed by expensive desktop software.
The reason is a decision made in 1993. The PDF specification defines fourteen standard fonts - Helvetica, Times, Courier, Symbol and Zapf Dingbats in their variants - that every reader is required to provide. A document using them does not embed them; it names them and relies on the reader. That was a sensible economy when a font file was a significant fraction of a floppy disk, and it is precisely what PDF/A forbids, because "every reader provides Helvetica" is a promise about the future that nobody is in a position to make.
When it finds a font that is named but not embedded, and that font is one of the standard fourteen text faces, it embeds a real outline for it. The substitutes are the Liberation family, published under the SIL Open Font License, and the choice is not aesthetic:
The face is subset to the characters the encoding can address rather than embedded whole, a character map is kept so the glyphs resolve, and a ToUnicode table is written so the archived document is still searchable and can still be copied out of. That last part is not required at level B and is done anyway, because a document that has preserved its appearance and thrown away its text has done half a job.
Three cases are reported rather than fixed, and each of them is a case where fixing it would mean guessing:
| Case | Why it is refused |
|---|---|
| A non-embedded font that is not one of the standard fourteen | A document naming Calibri or Frutiger without embedding it needs that specific font file. Substituting a different face would change the metrics, move every line break, and produce a file that conforms and no longer matches the original. |
| A font with a Differences encoding array | That array remaps character codes to glyph names, and honoring it needs the full Adobe glyph list. Substituting while ignoring the remapping produces a document whose text is confidently wrong rather than merely missing, which is the worse of the two failures. |
| Symbol and Zapf Dingbats | These are symbolic fonts with their own encodings and no Latin equivalent. There is no metrically compatible substitute to offer. |
In all three cases the result screen names the font and states plainly that the file will not pass a strict validator. That is deliberately uncomfortable, because the alternative - producing a file that says PDF/A on it and fails - moves the discovery of the problem to the worst possible moment.
The reliable fix for an unfixable font is upstream. Open the original document in whatever produced it and export again with "embed all fonts" or "embed fonts in file" switched on. Five minutes. Every word processor, layout application and print driver has that option, it is usually off by default, and it solves the problem at the only place where the font file actually exists.
A PDF that says "fill this rectangle with 0.9, 0.2, 0.2 in DeviceRGB" has not specified a color. It has specified three numbers and left the interpretation to whatever device renders them, which is exactly the kind of environmental dependency PDF/A exists to remove. The standard's answer is the output intent: an ICC color profile embedded in the file, describing precisely what its color numbers mean.
This converter writes an sRGB IEC 61966-2.1 profile, generated at conversion time rather than shipped as an asset. The colorants are the sRGB primaries already Bradford-adapted to the D50 profile connection space, which is what every conforming version of that profile carries, and the tone curves are sampled from the actual transfer function rather than from the 2.2 gamma approximation people quote. It is a real, parseable profile. A validator will open it and read it, because that is exactly what a validator does.
The standard requires an XMP packet declaring which part and level the file claims. Two things go wrong here. Both are fatal. It must be UTF-8 and it must be stored uncompressed, because a validator reads it without running a filter chain, so a deflated packet is an invisible one. And the values in it must agree with the document information dictionary: a file whose XMP title says one thing and whose Info title says another fails on that mismatch alone, regardless of everything else being perfect.
That is why this page asks for a title. Type something real. Whatever you type is written into both places, along with the author and subject if you supply them, so the two can never disagree. Leave it blank and the file name is used, which is better than an empty field and worse than a real title on a document somebody has to find in an archive in fifteen years.
PDF/A-1 is built on PDF 1.4, which predates both object streams and cross-reference streams. Writing those into a file that claims part 1 makes it unreadable by a conforming 1.4 reader and fails validation unconditionally. So when you select PDF/A-1b, this converter writes the older serialization, and when you select part 2 or 3 it uses the compact one. None of that is visible to you. It is worth about a fifth of the file size on a large document, and it is the kind of detail that separates a file which passes from one that merely looks like it should.
The tool offers two routes to the same standard, and the choice between them is the only real decision this page asks you to make.
Use rebuild when preserve has reported a font it cannot fix, you have no access to the original document, and the deadline is today. All three, not one of the three. It is a real answer to a real problem, and it is why the option is there. Do not use it as a default: an archive full of unsearchable page images is an archive nobody can use, and the loss is permanent in a way the original problem was not.
There is a third route that beats both and takes five minutes: go back to the source document and export it again with font embedding switched on. If you have the original file, that is always the correct fix, because it solves the problem where the font actually exists rather than working around it downstream. Try that first.
Think about which documents get converted to PDF/A. Tender submissions and signed contracts. Court filings. Medical records sitting under a statutory retention schedule. Board minutes, personnel files, financial statements that are not public yet. The set of documents that need archival conversion overlaps almost exactly with the set that must never be uploaded to a free web service to get it.
Every uploading converter asks you to make that trade, and most people make it because the alternative is buying a license for desktop software they will use twice a year. Not here. There is no upload endpoint, because there is no server-side step. Your file is opened with the browser File API, audited and rewritten by JavaScript on your processor, and the result is produced as a Blob your own browser downloads from itself.
One caveat, stated plainly. If a document needs a substitute font, that font file is downloaded from this site. It is a request for a public asset carrying no information about your document whatsoever, and it is the only network activity the conversion can cause. A document whose fonts are already embedded causes none at all.
This converter reports what it can check. It is not a validator, and you should not treat any tool's own report as proof of its own work. If the document is going somewhere that matters - a tender, a court, a national archive - validate it independently before you submit it, using something with no stake in the answer.
| Reported failure | What it means | Fix |
|---|---|---|
| Font not embedded | A face this converter could not substitute: not one of the standard fourteen, or carrying a Differences encoding. | Re-export the source document with font embedding switched on. Failing that, rebuild as images. |
| Transparency not permitted | You selected part 1 and the document uses transparency or a soft mask. | Select PDF/A-2b, which permits it. If part 1 is mandatory, rebuild as images. |
| LZW compression not permitted | A stream uses a legacy encoding the standard excludes. | Rebuild as images, or re-export from the source with modern compression. |
| External stream reference | Something in the file points at content outside it. | Rebuild as images. The reference cannot be resolved without the missing content. |
| Metadata does not match Info | The XMP packet and the document information dictionary disagree. | This converter writes both from one source, so a failure here means the file was edited afterwards. Convert again and do not touch it in between. |
| Structure tree missing | The validator was run at level A, not level B. | Level A cannot be produced from a flat PDF by any tool. The tagging has to be added by the software that created the document. |
Convert last, not first. PDF/A conversion should be the final step before archiving. Merging, splitting, rotating, watermarking or signing a file afterwards will usually break its conformance, because the tool doing it has no obligation to preserve the metadata or the output intent. Do all the editing, then convert, then validate, then submit.
Drop the PDF into the zone at the top of this page, choose which part of the standard you need, choose Preserve text or Rebuild as images, and press Convert to PDF/A. The conversion runs in your browser on your own processor, no account is needed for your first run, and the result screen tells you whether the file actually conforms rather than just claiming it does.
They are built on different versions of PDF and allow different things. PDF/A-1b is based on PDF 1.4 and forbids transparency, layers and attachments, so choose it only when a system asks for it by name. PDF/A-2b is based on PDF 1.7, allows transparency and layers, and is the current default for most archives and for European public procurement. PDF/A-3b is part 2 plus the right to carry file attachments, which is what electronic invoicing formats use to put machine readable invoice data inside a file a person can read. A higher number is not better. It is only newer.
No. No tool can produce it from a flat PDF. Level A also requires a full logical structure tree with tagged headings, reading order, alternative text on images and language markers, and none of that exists in an untagged PDF. It has to be added by the software that created the document, or by hand in a full editor. Any converter offering to turn an untagged PDF into level A is either inventing a structure tree or mislabeling the output.
Almost always because a font is named but not embedded and is not one of the standard fourteen that can be substituted with a metrically compatible face. Fonts like Calibri or a corporate typeface need that specific font file, and replacing them with something else would change every line break. The result screen names the fonts. The reliable fix is to re-export the source document with font embedding switched on, and the fallback is Rebuild as images.
No. The audit and the conversion run in your browser on your own processor, and your document is never transmitted anywhere. That matters more here than on most tools, because the documents people convert to PDF/A are tenders, contracts, court filings and records under a retention schedule. The one piece of network activity a conversion can cause is downloading a substitute font file from this site, which carries no information about your document, and a file whose fonts are already embedded causes none at all.
It substitutes, carefully. When the font is one of the standard fourteen text faces, it embeds a real outline from the Liberation family, which carries the same advance widths as the fonts it replaces: Liberation Sans matches Arial and therefore Helvetica, Liberation Serif matches Times New Roman, and Liberation Mono is fixed pitch like Courier. Where the document states its own width array, that array is kept untouched and the substitute simply supplies outlines for it, so no text can shift. A ToUnicode table is written as well, so the archived file stays searchable.
It should not. Substitution is metric preserving, so glyphs land on the coordinates the content stream already chose, and where the file carries its own width array nothing is recalculated at all. What can change visibly is a document that used a font the reader was previously substituting on the fly, since it now uses one specific face on every machine rather than a different guess on each - which is the entire point of the standard.
No, on two counts. Encrypted files cannot be parsed here, and PDF/A forbids encryption outright, so even if it could be opened the password would have to be removed for the file to conform. Remove the protection first with Unlock PDF, on a document you are authorized to open, then convert.
When Preserve text has reported a font it cannot fix, you have no access to the original document at all, and you need a conforming file today. Rasterizing guarantees conformance, because a document containing no fonts cannot violate the rules about fonts. The cost is permanent: the text cannot be searched, quoted or read by assistive technology, and the file is several times larger. Do not use it as a default.
Run it through veraPDF. That is the open source reference implementation maintained by the PDF Association, it is free, and it is what most archives actually run at intake, which makes it the only opinion that counts. Adobe Acrobat Pro has a PDF/A preflight profile that produces a readable report as well. Do not treat any converter's own report, including this one, as proof of its own work.
Because PDF/A requires the XMP metadata packet and the document information dictionary to agree with each other, and a mismatch between the two fails validation on its own regardless of everything else being correct. Two places, one value. Whatever you type is written into both, so they cannot disagree. Leave it blank and the file name is used, which is better than an empty field and worse than a real title on something somebody has to find in an archive in fifteen years.
You can. It will usually stop conforming. Merging, splitting, rotating, watermarking or signing a file afterwards is done by tools that have no obligation to preserve the output intent or the metadata packet. Treat PDF/A conversion as the final step: do all the editing first, then convert, then validate, then submit.
Yes. 20 files and 100 MB per batch for anonymous visitors and free accounts, and 100 files and 500 MB on CartHacker Premium. Each document is audited and converted separately and gets its own entry in the report, so a batch where three files conform and one does not tells you exactly which one to look at rather than giving you a single verdict for the lot. Anonymous visitors get one operation per 12 hours, a free account gets two, and Premium removes the counter.
All of these run the same way: your file is read by JavaScript on your own machine, the work happens on your processor, and the result is handed back to you as a download that never existed anywhere else.
Find the apps that slow your store down, by name and in milliseconds, then answer shoppers' doubts with trust badges under Add to cart. Both are free, with no card and no access to your store admin.
Paste any page and get its three Core Web Vitals, the apps costing time, like Klaviyo or Judge.me, and the Shopify or WooCommerce fix.
Show the promises that settle doubts, from free shipping to a money-back guarantee, right under the Add to cart button. One line in your theme.
Change the link after printing and count the scans. 5 tracked codes on a free account, static codes unlimited.
Anonymous visitors get one operation per 12 hours, per network address. That is not a sales tactic, it is the only lever we have: this tool runs on your processor, not ours, so the run itself earns us nothing. An account tells us who you are, and the ceiling goes up the moment you have one. Your allowance resets in 12 h.
No credit card, no trial period, no sales call. Join with Google, GitHub, Facebook or email.
Free accounts get two operations per 12 hours, and yours reset in 12 h. Waiting costs you nothing and is a completely reasonable choice. If PDF work is a weekly job rather than a one-off, the counter is the wrong tool for you and Premium takes it off the page.
Cancel anytime. Everything you have already produced stays yours either way.