Why does my ebook have no table of contents?
You open the contents menu on your e-reader and it's empty, or it lists file names such as text/ch07.xhtml where the chapters should be, even though the book has a contents page near the front. That page is only text with links. The menu is filled from a separate, machine-readable table of contents inside the file, and your book either lacks one or has the kind your reader ignores.
An EPUB can carry three tables of contents. One is the page you can read. Another is the NCX, an XML file that the EPUB 2 package specification made compulsory: an EPUB 2 book lists it in its manifest and names it in the toc attribute of the spine. The third is the EPUB 3 navigation document, an XHTML file holding a <nav epub:type="toc"> list of links. EPUB 3.3 requires exactly one such list in every book and files the NCX under legacy features, there for EPUB 2 reading systems. The reading system specification obliges an EPUB 3 reader to show you its links.
Reading apps differ in what they do when that list is missing. Kobo spells out its rules in the EPUB guidelines it keeps on GitHub, which I read in November 2026: for an EPUB 2 the menu comes from the NCX, and with no NCX a Kobo lists the files of the spine; a file the NCX leaves out gets an entry named after the file or the opening words of the section. For an EPUB 3, Kobo reads the toc nav alone and ignores any NCX. With no toc nav, Kobo e-readers and the desktop app fall back to file names, the iOS app to the page list, and the Android app to the page list or the landmarks. If the Android app finds neither, its menu is empty.
Amazon insists on the machine-readable kind. Its KDP help page on the table of contents (November 2026) calls a logical table of contents, built from a toc nav or an NCX, “Required for all Kindle books by Amazon”, because it fills the Go To menu. The visible contents page is only strongly recommended.
How a book loses it
A converter can keep only a table of contents that it knows how to read, and the file you start from may not have one at all. calibre's conversion manual describes the fallback: without a table of contents in the input's metadata, calibre detects chapters (by default, h1 and h2 headings containing words such as “chapter” or “part”, or elements with the class chapter) and adds links from the text when it finds fewer chapters than a set threshold. Chapter titles set as plain bold, centred paragraphs match none of those rules, so the menu comes out thin or empty. Readwyn's PDF conversion takes the contents from the PDF's outline, or from its headings when the PDF has no outline.
An EPUB can be repaired. calibre's editor has Tools › Table of Contents › Edit Table of Contents, which its manual says can generate entries from the headings, from links, or from the individual files, and lets you aim any entry by clicking in a preview.