The format question eventually catches up with anyone who reads e-books outside a single ecosystem. A Tolino wants EPUB. A Kindle, on the other hand, speaks AZW3 or KFX. MOBI files from the 2000s are still floating around online, and the bookshop delivers a DRM-protected EPUB that simply refuses to go onto a Kindle. How the formats are built, which device handles what, why Amazon retired MOBI, and where the real limits of conversion lie – all of this is explained in this article.
What an e-book actually is
The same basic construction is shared by almost every common e-book format: it is a container – an archive – bundling text files in some flavour of HTML together with stylesheets, images, fonts and a metadata file. The reader unpacks the container, renders the HTML and reflows the text to the current screen size.
The key difference from PDF lies in that reflow. In the narrow sense, an e-book is reflowable: it has no fixed page count and adapts to display size, font size and margins. A PDF, by contrast, is fixed layout. Every page has set dimensions, and on a 6-inch display you end up scrolling and zooming. Converting PDF to EPUB is therefore technically demanding, while EPUB to PDF is almost always clean – a layout is simply being imposed where none existed before.
The language inside the container is where the second important difference lies. EPUB 3 uses XHTML5 and a large part of CSS. MOBI, however, uses a subset of HTML 3.2 from the 1990s. For that reason, layout is inevitably lost when a modern EPUB is converted to MOBI – not because the converter is bad, but because the target format has no concept of those constructs.
EPUB: the open standard
Of the formats covered here, EPUB is the only one that is a genuine open standard. The International Digital Publishing Forum (IDPF) developed it. That organisation merged into the W3C in 2017, and the standard has been maintained there ever since. EPUB 3.3 has been a W3C Recommendation since May 2023.
Technically, an EPUB file is a ZIP archive. Its structure is strictly prescribed:
- A
mimetypefile. It must be the first entry in the archive, and it must be stored uncompressed – this is how readers identify the format reliably. META-INF/container.xml, which points to the package file.- An OPF file. It holds metadata (title, author, language, ISBN), a manifest of every resource and the reading order (the spine).
- The content itself as XHTML files, plus CSS, images and optionally embedded fonts.
- A navigation document (EPUB 3) or an NCX file (EPUB 2) for the table of contents.
An EPUB file can simply be renamed to .zip, and you can look inside. That is also why EPUB works so well as an interchange format: it is inspectable, validatable (with the official EPUBCheck tool) and readable in the long term.
The version history in brief: EPUB 2 (2007, final revision 2.0.1 in 2010) remains the lowest common denominator. It runs practically everywhere. EPUB 3 (2011) brought XHTML5 and considerably more CSS. MathML, SVG, embedded audio and video, JavaScript, media overlays for read-aloud features and fixed-layout support for comics and picture books were added as well. Versions 3.2 (2019) and 3.3 (2023) mainly tidied up; in addition, they aligned the standard with the web platform.
Another practical factor is accessibility. EPUB Accessibility describes how an e-book becomes usable with screen readers – through semantic markup, alternative text and a sound reading order. Since 28 June 2025, the European Accessibility Act has applied, and it explicitly covers e-books. For publishers, EPUB is therefore not only the technically but also the legally safe choice.
MOBI: the format Amazon inherited and then buried
MOBI comes from the French company Mobipocket; Amazon acquired that company in 2005. The format is considerably older than EPUB – and that shows everywhere.
The container is a Palm Database (PDB) – the same base format that Palm handhelds used in the late 1990s. Inside, the text sits as a subset of HTML 3.2. It is compressed with PalmDOC LZ77 or optionally with a Huffman scheme. From that follow the well-known limitations:
- No real CSS. Formatting runs through HTML attributes. Margins, font families or multi-column layouts therefore cannot be controlled reliably.
- No embedded fonts. The reader uses its own.
- JPEG and GIF only for images, historically with a very low size ceiling per image.
- No fixed layout – comics or picture books cannot be represented sensibly.
- No nested table structures, no SVG, no MathML.
MOBI was retired by Amazon in several steps. On 28 June 2021, Kindle Direct Publishing stopped accepting MOBI files for new reflowable titles. In August 2022, Amazon ended MOBI delivery through "Send to Kindle"; EPUB has been accepted instead since then. So if you hold a MOBI file today, you hold an archival format. It remains readable on older devices and in software such as Calibre, but it has no future in the Amazon ecosystem. The sensible response is to move it into an open format, for example with MOBI → EPUB.
AZW, AZW3 and KFX: Amazon's own ladder
Over the years, Amazon has stacked up a whole series of proprietary formats. They are easy to confuse.
AZW
In essence, AZW is a MOBI file with Amazon's DRM. Its technical capabilities and limits match MOBI. The difference lies in the encryption and in the binding to a customer account.
AZW3 (Kindle Format 8)
KF8 arrived in 2011 with the Kindle Fire, carrying the .azw3 extension, and reached the e-ink devices in 2012. This format is the big step forward: KF8 supports a subset of HTML5 and a substantial part of CSS3. Embedded fonts, drop caps, nested tables, numbered lists, SVG and fixed-layout content are thereby made possible. In effect, AZW3 is Amazon's answer to EPUB 3 – roughly comparable in capability, but proprietary.
One curiosity from that transition period: Amazon's Kindlegen tool of the time frequently produced dual-format files with a .mobi extension. Internally, these contained both an old MOBI section and a KF8 section. New devices read the KF8 part, old ones the MOBI part – at the cost of a file roughly twice as large as necessary.
KFX
Since around 2015, KFX has been Amazon's current format. It forms the basis for so-called Enhanced Typesetting: improved hyphenation, ligatures, kerning, cleaner justification, plus features such as Page Flip and Word Wise. Compared with AZW3, KFX is considerably more locked down. It is normally not sideloaded but generated server-side by Amazon. Today, when you send an EPUB via Send to Kindle, Amazon converts it to KFX on its servers.
Head-to-head comparison
| Property | EPUB 3 | MOBI | AZW3 (KF8) | KFX |
|---|---|---|---|---|
| Introduced | 2011 | 2000s | 2011 | approx. 2015 |
| Standardised | W3C, open | proprietary | proprietary | proprietary |
| Container | ZIP | Palm Database | Palm Database (KF8) | own binary format |
| Markup | XHTML5 | HTML 3.2 subset | HTML5 subset | internal, generated from HTML |
| CSS | extensive | practically none | CSS3 subset | yes, plus its own engine |
| Embedded fonts | yes | no | yes | yes |
| Fixed layout | yes | no | yes | yes |
| Audio/video | yes | no | limited | limited |
| Accessibility | dedicated standard | barely any | limited | limited |
| Status | current | discontinued | legacy/sideload | current at Amazon |
Which device reads what?
| Device / app | EPUB | MOBI | AZW3 | Note |
|---|---|---|---|---|
| Kindle (e-ink) | only via Send to Kindle (converted to KFX) | sideloaded only, older devices | yes, over USB | No direct EPUB sideloading over USB |
| Tolino | yes | no | no | EPUB plus Adobe DRM |
| Kobo | yes | yes | no | Uses KEPUB internally, its own EPUB variant |
| PocketBook | yes | yes | partly | Very broad format support |
| Apple Books | yes | no | no | EPUB and PDF |
| Google Play Books | yes | no | no | EPUB and PDF |
| Calibre | yes | yes | yes | Reads and writes practically anything without DRM |
The important row is the first one. EPUB will not be accepted by a Kindle over a USB cable. The route runs through Send to Kindle – by email or app – with Amazon converting the file server-side. If you would rather avoid that, or you have an older device, convert beforehand yourself, for instance with EPUB → MOBI for older Kindles.
The reverse also holds: with AZW3 → EPUB, an AZW3 from your Kindle library can be moved to a Tolino or Kobo – provided the file is free of DRM.
DRM: the real reason conversions fail
Most everyday format problems are not format problems at all. They are copy-protection problems. Digital rights management comes in two large camps:
- Amazon DRM is found in AZW, AZW3 and KFX. The file is tied to a customer account and sometimes to a device ID.
- Adobe DRM (ADEPT) and increasingly Readium LCP are found in EPUB files from the rest of the trade. Here the file is unlocked with an Adobe ID or a passphrase.
A DRM-protected e-book is encrypted. No converter can do anything with it, because the content simply cannot be read – typically the conversion aborts with an error or produces an empty file. That is not a bug. It is the point of the encryption. Whether and how DRM may be removed is a matter of copyright law. In Germany, circumventing effective technical protection measures is generally prohibited under section 95a of the Copyright Act. Comparable rules exist across the EU. The cleanest route, if you want to read your books independently of format, is to buy them DRM-free in the first place – many technical publishers and platforms offer exactly that.
A practical test: the EPUB is DRM-free if it can be renamed to .zip and unpacked normally. If, however, you find a META-INF/encryption.xml or rights.xml inside, it is protected.
What conversion actually does
A converter reads the container and interprets the markup inside; it then writes that markup into the target structure. As with any format transformation, the same rule applies: whatever the target format does not know is lost. Concretely:
- EPUB → MOBI: embedded fonts, elaborate CSS, SVG and MathML fall away. Text, chapter structure and table of contents, however, survive.
- MOBI → EPUB: low loss, because the target format can do more than the source. But whatever was never in the MOBI does not magically appear either.
- AZW3 → EPUB: usually very clean, since both formats build on HTML/CSS and offer a similar feature set.
- EPUB → PDF: reflowable text becomes a fixed layout. This is useful for printing or archiving, but the way back is laborious.
Metadata is a frequent stumbling block. Title, author and language live in the OPF file. When an e-book comes from a poor source, those fields are often empty or wrong. After conversion, the library then files the book under "Unknown". In Calibre or an EPUB editor, this can be fixed in a minute – you just need to know where to look.
If you manage books from the Russian-speaking world, FB2 will also frequently cross your path – an XML-based format with its own structure. A direct route exists for that too, with FB2 → EPUB.
Recommendation: what to use when
| Situation | Format | Reasoning |
|---|---|---|
| Archiving your own library | EPUB (DRM-free where possible) | Open standard, inspectable, readable long term |
| Current Kindle | EPUB via Send to Kindle | Amazon turns it into KFX with the best typography |
| Older Kindle without Send to Kindle | AZW3, MOBI as fallback | Transferable directly over USB |
| Tolino, Kobo, PocketBook | EPUB | The native format of these devices |
| Comics, picture books, textbooks | EPUB 3 fixed layout or PDF | The layout has to survive |
| Printing or official submission | Fixed pages, defined rendering |
Frequently asked questions
Can I just copy an EPUB onto my Kindle over USB?
No. Sideloaded EPUB files are not read directly by Kindle devices. Either use Send to Kindle, which has Amazon convert the file server-side, or convert it yourself beforehand to AZW3 or MOBI and transfer the result over the cable.
Is AZW3 better than EPUB?
Functionally, the two are roughly on par – AZW3 was Amazon's counterpart to EPUB 3. The difference is strategic. EPUB is an open W3C standard that many devices read, whereas AZW3 works effectively only inside the Kindle world. For your own collection, EPUB is therefore the more robust choice.
Why does my conversion abort with an error?
In the vast majority of cases, DRM is the reason. A protected file is encrypted and unreadable to the converter. A quick test: rename the EPUB to .zip and unpack it. If that fails, or there is an encryption.xml inside, the file is protected.
Do I lose formatting when converting?
It depends on the direction. Practically nothing is lost when going from the poorer to the richer format – MOBI to EPUB, say. In the other direction, elements the target format does not know fall away: embedded fonts, complex CSS, SVG. Text and chapter structure survive in both directions.
What about KFX – can I convert that?
KFX is Amazon's current, heavily locked-down format. It is practically always DRM-protected and is not intended as an interchange format. If you want to keep your books format-independent, acquire them DRM-free and archive them as EPUB rather than trying to convert out of KFX later.
Conclusion
Tidier than it looks – that is the e-book format landscape. MOBI is a relic of the Palm era; Amazon finally retired it in 2021 and 2022. It is only useful for old devices now. AZW3 was the technically solid intermediate step; today it matters mainly for sideloading onto Kindle devices. KFX is Amazon's present, but a closed end format, not an interchange format. That leaves EPUB as the only open, standardised and broadly supported option. The rule for your own collection follows directly: buy DRM-free, archive as EPUB, and derive from it whatever the device in your hand happens to need.
Convert e-books with wandlio
These converters match the routes described here:
- MOBI → EPUB: move legacy files into an open format
- AZW3 → EPUB: make Kindle books usable on Tolino, Kobo and PocketBook
- EPUB → MOBI: for older Kindle devices without Send to Kindle
- EPUB → PDF: a fixed layout for printing or archiving
- FB2 → EPUB: bring FictionBook files into the standard
Document conversions run server-side; after the conversion, your file is deleted. DRM-protected files cannot be converted.