Accuracy: How This Generator Is Tested
A citation generator is a correctness product: one wrong comma costs a student marks. This page publishes how the formatter is tested, what it is tested against, and the results — and it is updated every time the test file is re-run.
The method
The formatter is checked against a fixed test file of 36 official-example cases: the 18 combinations of style and source type this site generates (APA, MLA and Chicago × website, journal article, book, book chapter, online video, newspaper article), with two cases each, taken from or modelled directly on the official examples in the sources of truth below.
The cases deliberately cover the awkward records, not just the tidy ones: one, two, three or more and twenty-one or more authors; a corporate author; no author; no date; a chapter with two editors; a book with an edition; a journal article with a DOI and one without; a website with and without an author; a video whose uploader is not the author; and newspaper articles online and in print.
A case passes only when the generated entry matches the expected format character for character — italics, punctuation, capitalisation and spacing included. Every mismatch is logged, the formatter is fixed, and the whole file is re-run. Every case below is re-run after any change to the formatter. Lookup (autocite) output passes through the same expectations: convenience never bypasses the format gate.
Named failure-mode regressions
Beyond the 36 cases, the test file carries named regression cases for failure modes seen in the field, so a fixed bug can never quietly come back:
| Regression case | What is asserted |
|---|---|
| Duplicate DOI | A DOI entered in every messy form — doi:10.…, an old dx.doi.org address, a bare DOI, and an already-prefixed https://doi.org/… URL — always produces exactly one https://doi.org/… in the output, with no full stop after it. |
| Sentence case | An article title typed in title case comes out in APA sentence case. A proper noun typed in lowercase is flagged for the student’s own check rather than silently “fixed”. |
| Wrong-record match | Where a search returns a same-title record for a different source, the resolver either resolves the exact identifier given or stops at a confirmation step — it never formats the wrong source. |
| ISBN lookup | A book resolved by ISBN through Open Library / Google Books keeps its edition and publisher intact in the generated entry. |
Sources of truth
- APA: the APA Style site (apastyle.apa.org), its style-and-grammar guidelines and reference examples, checked against the Publication Manual of the American Psychological Association, 7th edition.
- MLA: the MLA Handbook, 9th edition (the core-elements model), and the MLA Style Center (style.mla.org) Works Cited guidance.
- Chicago: The Chicago Manual of Style, 18th edition (2024), chapters 13–14 documentation guidance and the CMOS quick-guide examples.
No template on this site is coded from a competitor’s output. Competitor pages are used, at most, as a cross-check alarm that sends us back to the manuals.
Results
| Test group | Cases | Result |
|---|---|---|
| Official examples — APA 7th edition | 12 | 12 / 12 passed. |
| Official examples — MLA 9th edition | 12 | 12 / 12 passed. |
| Official examples — Chicago 18th edition | 12 | 12 / 12 passed. |
| Failure-mode regressions (duplicate DOI, sentence case, wrong-record match, ISBN lookup) | 4 named cases | Passed in full — 7 / 7 automated format checks (the duplicate-DOI case in all four messy forms, APA sentence case, MLA title case kept, and the ISBN case’s resolved record), and the behaviour cases (lowercase proper-noun flag, wrong-record match, ISBN lookup) verified end-to-end in a live browser, 11 / 11 sub-checks. |
| Overall | 36 cases + regressions | 43 / 43 format checks passed (36 official-example cases + 7 regression checks); 3 / 3 behaviour cases verified (11 / 11 sub-checks); pre-launch layout audit 68 / 68 page-views with no problems. Last full run: 21 October 2026. |
Claims on this page never go beyond what the current test file shows. If a style cell cannot yet be verified against an official example, the page says so rather than guessing.
One honest limit applies to every generator, including this one:
Check your reference against your institution’s guidelines — lecturers sometimes vary the style.