A PDF states, in its first line, which version of the specification it was written against. We read that line on every PDF delivered by a live listing — 2,630 files, measured 19 September 2026 — and nearly six in ten of them name a specification published in 2001.
That would be a curiosity if the version number were the whole story. It is not. The same parse shows 59.32% of these documents embed no fonts at all, and that is the part a buyer notices.
What the declared version says
Declared PDF version across the 2,630 files:
- 1.4 — 1,553 files (59.05%)
- 1.7 — 738 files (28.06%)
- 1.3 — 182 files (6.92%)
- 1.5 — 91 files
- 1.6 — 64 files
- 1.2 — 1 file
- 2.0 — 1 file
PDF 1.4 is the 2001 specification. PDF 1.7 is the 2006 revision that became ISO 32000-1, and it is what most software treats as today's baseline. PDF 2.0 has existed since 2017, and exactly one document on the shelf is written to it.
Nobody sat down and picked 1.4. It is what the export defaults of common tools still choose, and those defaults were tuned for a question — will a reader from a few years back open this? — that stopped mattering a decade ago. The version number on its own costs you very little. What travels with it does not.
Six in ten documents carry no embedded fonts
Of the same 2,630 files:
- 59.32% embed zero fonts
- 52.62% do not embed every font they use
- Median fonts per document: 2. 90th percentile: 20. Maximum: 3,888
The first two figures overlap rather than contradict. A file can embed nothing because it references only standard faces, and a file can embed some of what it needs while leaving the rest to chance. The second case is the dangerous one, because it looks finished.
When a font is missing, the reader substitutes something it already has. Text still appears, so nothing looks broken to the person who exported it — their machine has the font. On a device that does not, the substitute has different metrics. Line breaks move. A two-page spread becomes three. A table stops lining up. In non-Latin scripts, glyphs can fall back to empty boxes. Phones are the worst case, because they carry the fewest faces.
The fix is one checkbox. In Canva, Figma, Word, Google Docs or InDesign, look for "embed fonts" in the PDF export dialog, or export as PDF/A, which requires embedding by definition. Then verify it the only way that proves anything: open the file on a device that does not have your fonts installed.
The 7.41% with nothing to select
92.59% of the shelf carries a text layer. The remaining 7.41% is image-only — either scanned, or exported as pictures of pages rather than as pages.
An image-only PDF cannot be searched, cannot be copied from, and is invisible to a screen reader. For a one-page worksheet that is an annoyance. For a long reference document it removes the feature that made it a reference, and it means nobody can paste a line of it into anything — often the exact reason they wanted it.
If your source is a scan, run optical character recognition before export; most scanner apps and Acrobat do it in one pass. If your source is a design tool, check that you exported the document rather than a render of it. The setting is usually called "flatten" or "rasterize", and turning it off is the whole fix.
One smaller number from the same parse: 13 files, 0.49%, are encrypted. Encryption on a paid download is not protection — the password has to travel with the file — but it is reliably a support question, because some readers on some phones decline to open it at all.
What to take from this
None of this is about making better work. It is about whether the work arrives looking the way you built it. Three checks, in the order they pay off:
Embed the fonts. One setting decides whether your layout survives the trip.
Keep the text as text. Selectable, searchable, readable aloud.
Open it somewhere else. A different machine, or a phone. Five minutes, once, and it catches every one of the above.
If you list documents, saying which of these is true is worth a line in the description. Listings that name the format precisely get asked fewer questions afterwards.



