What PDF/A actually is
PDF/A is an ISO standard (19005) that takes ordinary PDF and removes everything that depends on the outside world or on a particular viewer. Fonts must be embedded, because the reader in 2050 will not have your fonts. Colour must be defined by an embedded ICC profile, not “whatever the device does”. Encryption, JavaScript, external links to content, embedded files (at level 1 and 2), audio and video are forbidden, because they break, expire or require software. XMP metadata must describe the file and declare its conformance. In return, an archive can accept the file knowing it is self-contained.
Who asks for it
- Courts and e-filing systems in many jurisdictions require PDF/A for filings — often 1b or 2b. See PDF for court e-filing.
- Land, company and patent registries, and tax authorities that accept electronic records.
- Universities for theses and dissertations.
- Public archives and records offices, and any organisation with a retention policy measured in decades.
- Accounting and audit, where invoices and records must remain readable for statutory periods.
Which level
| Level | What it adds | When it is asked for |
|---|---|---|
| PDF/A-1b | The original (2005) profile, based on PDF 1.4; no transparency, no JPEG 2000 | Older requirements; usually satisfied by 2b as well |
| PDF/A-2b | Based on PDF 1.7; allows transparency, JPEG 2000, layers, embedded PDF/A attachments | The common requirement for “visually preserved” documents |
| PDF/A-2a | 2b plus tagged structure and Unicode mapping for accessibility | Accessibility mandates, public bodies |
| PDF/A-2u | 2b plus Unicode mapping for all text | Searchable archives without full tagging |
| PDF/A-3 | 2 plus arbitrary embedded files (for example the XML inside an e-invoice) | Electronic invoicing standards |
If a requirement says “PDF/A” without a level, 2b is the safe answer: it satisfies almost every 1b requirement in practice and allows modern content. Level a needs tags that only the authoring tool can produce properly.
Converting on your device
PDF to PDF/A runs a PDF/A-2b preflight first — fonts embedded, encryption, JavaScript, embedded files, multimedia annotations, annotation appearances, output intent, XMP, document ID — and shows every objection. Conversion then embeds an sRGB ICC output intent, declares PDF/A-2b in XMP, sets a document ID, removes forbidden content and generates form appearances. The same preflight runs again on the result, and the report downloads with the file. Nothing is uploaded, which matters for the court filings and personal records that need this most.
The font problem, honestly
Text set in a font that was never embedded is the objection converters hide. Without the font file, the text cannot be re-encoded, so the only conforming option is to draw the page as an image — Bindery does this at 200 DPI, tells you which pages, and lets you decline and keep the text with the objection recorded. The better fix, when possible, is to re-export from the authoring tool with all fonts embedded and convert that.
Checking the result
No browser tool can certify conformance; validators such as veraPDF do that, and archives run them. Bindery’s report lists every rule it covered so a formal check holds no surprises. Before converting, give scans a text layer with OCR, set a title and language with Remove metadata, and if level a is on the horizon, run the accessibility check to see what tagging is missing. Convert last: any later edit can break conformance.