Ask three suppliers for PDFs and you can receive three genuinely different things. One sends a photograph of each page. One sends the same photograph with searchable text hidden behind it. One sends a file built to open correctly in twenty years. They all end in .pdf, and only one of them may be what your retention policy actually requires.
Three files that look identical on screen
| Image-only PDF | Searchable PDF | PDF/A | |
|---|---|---|---|
| What is inside | Page images only | Page images plus an invisible OCR text layer | Page images, text layer, and every font, colour profile and resource the file needs |
| Can you search it? | No | Yes, as far as the OCR read correctly | Yes, where a text layer was added before archiving |
| Can you copy text out? | No | Yes | Yes |
| Appearance | The scanned page, unchanged | The scanned page, unchanged | The scanned page, unchanged, and guaranteed to render the same way later |
| External dependencies | None, but nothing to depend on either | May rely on fonts installed on the reader machine | None permitted. Fonts and profiles are embedded |
| Encryption and scripting | Permitted | Permitted | Not permitted, which is the point: nothing can block future access |
| Typical use | Casual reference copies | Day-to-day operational records | Statutory, legal and long-retention archives |
| File size | Smallest | Marginally larger than image-only | Largest, because everything travels inside the file |
How the text layer works
In a searchable PDF the recognised text is written into the file at zero opacity, positioned over the words in the image. Your eye sees the scan. Your search sees the text. Selecting a paragraph highlights it in the right place because the invisible characters sit where the visible ones are.
Keeping the image is deliberate. For a signed contract, a stamped approval or anything that may be produced as evidence, the picture of the page is the record. Replacing it with reconstructed text would discard the signature, the letterhead, the ink of the stamp and the visible fact that nothing was altered.
What PDF/A is actually for
PDF/A is an ISO-standardised subset of PDF designed for preservation. It forbids the features that make a normal PDF fragile over time. No linked external fonts, so the file cannot render in a substitute typeface once the original font is gone. No encryption, so nobody in 2045 is locked out by a password nobody recorded. No embedded scripting or external content. Everything the file needs to display is inside the file.
The conformance levels differ in what they permit and what they promise. PDF/A-1 is the strictest and the most widely supported. Later parts allow useful additions, including embedding a source data file alongside the visual document, which is how a structured invoice can travel inside its own human-readable rendering. Which part you want depends on your retention rules and the systems that will hold the archive, and it is a question to settle before capture rather than after.
One honest limitation: PDF/A guarantees that the file will render as intended. It does not guarantee that the text layer inside it is correct. Those are separate concerns, and only the first is a property of the format.
Compression, file size and small print
Compression settings are where archives quietly lose information. Aggressive segmentation of text from background produces very small files and looks fine at normal zoom, but can smear the small print, footnotes and Arabic diacritics that somebody will eventually need to read closely. Bitonal conversion of a faint original can erase the faintest strokes altogether.
We set compression per document class rather than per project. Reference copies of routine correspondence can be compressed hard. Contracts, technical drawings and anything with dense small print are treated conservatively, on the basis that storage is cheaper than a rescan of paper that may no longer exist.
Choosing between formats
- Operational records people search daily: searchable PDF, compressed sensibly, indexed for retrieval.
- Statutory, legal and long-retention records: PDF/A at the conformance level your policy specifies, with the text layer added before conversion.
- Records feeding another system rather than a reader: searchable PDF for the human copy, plus a structured export for the machine.
- Existing image-only archives: recognition can be applied and the files converted in place, so you are not rescanning paper you already handled once.
Delivered files are validated against the target format before handover, and the index values can be embedded in the file metadata as well as supplied in the load file. That way a document separated from its repository still carries some idea of what it is.
Frequently asked questions
What is a searchable PDF?
A searchable PDF contains the scanned page image plus an invisible text layer produced by OCR, positioned behind the visible words. The document looks exactly like the original scan but can be searched, copied and indexed by content. It is the standard deliverable for records people need to read and find.
How do I tell whether a PDF is searchable?
Open it and try to select a line of text with the cursor, or run a find for a word you can see on the page. If nothing selects and nothing is found, the file is image-only. Text selecting slightly out of position usually means a text layer is present but was placed by OCR.
What is PDF/A and do we need it?
PDF/A is an ISO archival subset of PDF that embeds fonts and colour profiles and forbids encryption and external dependencies, so the file renders the same way decades later. It is worth using for statutory, legal and long-retention records. For records superseded within a few years, a standard searchable PDF is usually adequate.
Can you convert our existing image-only PDFs?
Yes. Recognition can be run across an existing image archive and the files rewritten as searchable PDF or PDF/A without touching the original paper. The quality of the resulting text layer depends on how those images were captured, particularly their resolution and compression.
Does adding a text layer change how the document looks?
No. The recognised characters are written at zero opacity behind the page image, so the visible document is still the scan, including signatures, stamps and letterhead. This matters where the image itself is the evidential record and reconstructed text would not be acceptable.
Why are PDF/A files larger than our current PDFs?
Because everything needed to display the file travels inside it, including embedded fonts and colour profiles, and because archival conversion generally uses more conservative compression. The size difference is the cost of not depending on software or resources that may not exist when the file is next opened.
Related reading
- OCR and text conversionThe recognition step that produces the text layer inside a searchable PDF.
- Document indexing and metadataIndex fields find the document; the text layer searches inside it.
- Retention and complianceWhich of your records justify archival-format conversion.
- Glossary: PDF/AShort definition of the archival format and its conformance levels.