How accurately IndexerLabs builds an index of ancient sources (Stellenregister): tested on ten published Mohr Siebeck books in English and German, checked entry by entry against the printed pages.
We took ten books that Mohr Siebeck published with an index of ancient sources made by hand: books on the Bible, the Dead Sea Scrolls, Philo, Josephus, rabbinic literature, papyri and inscriptions. The engine indexed every page. A separate reviewer then read the pages against their scans and checked each entry the engine made: is it really a citation printed on that page, of that work, at that passage? It also listed every citation the page prints that the engine left out.
To keep the test honest, the reviewer checked every page where earlier versions of the engine had gone wrong, plus a random tenth of all the other pages. The figures for whole books combine the two, so they are neither inflated nor flattered by the choice of pages. Measured October 2026.
| Book | Language | Pages | Entries correct | Citations found | See it |
|---|---|---|---|---|---|
| Feldmeier, Niehoff (eds.), Abrahams Aufbruch | German | 264 | 99.8% | 99.9% | Editor · Compare |
| Runesson, Judaism for Gentiles | English | 342 | 99.7% | > 99.9% | Editor · Compare |
| Stettler, Das Endgericht bei Paulus | German | 338 | 99.6% | 99.6% | Editor · Compare |
| Burz-Tropper (ed.), Studien zum Gottesbild im Johannesevangelium | German | 278 | 99.6% | 99.4% | Editor · Compare |
| Heilig (ed.), Empire Criticism of the New Testament | English | 372 | 99.5% | 99.6% | Editor · Compare |
| Stettler, Das letzte Gericht | German | 272 | 99.1% | 99.5% | Editor · Compare |
| Bühner, Paulus im Kontext des Diasporajudentums | German | 384 | 98.8% | 99.7% | Editor · Compare |
| Witte, Schröter, Lepper (eds.), Torah, Temple, Land | English | 300 | 98.4% | 99.9% | Editor · Compare |
| Brodersen, The Beginning of the Biblical Canon and Ben Sira | English | 192 | 98.3% | 97.7% | Editor · Compare |
| Zellentin, Rabbinic Parodies of Jewish and Christian Literature | English | 236 | 98.2% | 99.4% | Editor · Compare |
| All ten books | 2,978 | 99.2% | 99.4% |
Each book's figures rest on a sample of its pages and can move by about a point either way; the totals rest on all 761 pages checked. "Editor" opens the book and its index in our editor, read-only; "Compare" lists every citation with the pages the printed index gives beside the pages IndexerLabs gives.
Because the engine reads every page, it finds where a published index does not match its own book, and where a book prints a reference that cannot exist.
Each book's comparison page lists every mistake found in its printed index.
The engine finds 99.2% of the citations that the books' own published indexes record on the page. It also indexes many citations those indexes leave out: passages in footnotes, items deep in a list, references to papyri and inscriptions.
Along the way it surfaced dozens of places where a published index does not match its own book: a passage filed under the wrong biblical book, a verse number mistyped, a section number dropped from a list. An index typed by hand copies these forward; the engine reads the page.
Fewer than one entry in a hundred is wrong and fewer than one citation in a hundred is missed. What is left is mostly not a matter of reading the page more carefully.
Some indexes list a work named without a passage ("Jubilees retells Genesis"), or a Dead Sea Scroll named only by its siglum ("1QS"); others leave them out. Some list every verse number inside a passage the book reprints; others list the passage once. No single rule matches every book. Whether to list works named without a passage is a setting you choose at upload; for the others the engine follows the most common practice.
Authors and copy editors sometimes print a reference that cannot exist, such as a verse past the end of its chapter. The engine keeps the reference exactly as printed, because guessing the intended passage would be worse, and flags it for review.
A bare verse number that takes its book and chapter from earlier in the paragraph ("v. 12"), the last item of a long list in a footnote, a cell deep in a table of verses. These are where most of the remaining misses sit. Citations split across a page break and fragments of ancient authors known only through the writer who quotes them, which used to be missed, are now caught.
Where the page itself cannot settle a reference, such as an author's typo, the engine flags it for review rather than guessing.
An index of ancient sources in English or German, sectioned the way scholarly indexes are, for $199 per book.
See pricing