AI code detectors ranked by the programming languages the vendor names

Field sorted: the code language count field on each card, written on 9 November 2026 for the cards whose modalities list code, from the input limits and product fields read on 9 November 2026 and, for Originality.ai, from the 5.0 release post re-read live on 9 November 2026 because the card held no language list for code. A language counts when the vendor names it on a page recorded on the card as a language its code detection accepts or was tested on; a count the vendor states without names ("65+", "38+ more") is printed and not counted; a name on an archived page that the live page no longer carries is printed and not counted (Codeleaks' six). A card qualifies when it states code as a modality and names at least one language; a card that states code and names none (Copyleaks) is tabled apart; a card with no code modality is listed with the modalities it does state. Direction: more languages rank higher. Tie rule: more study records first, then product name; no tie arises among the three. 3 of 72 qualify. Full rules: methodology page. The page says which vendor names the most languages and never which code detector works best, because no indexed study has scored one.

The three vendors name languages for three different reasons. Codequiry's list is an intake statement, the file types the checker accepts, so it runs to twelve names and a headline of 65+ (codequiry.com, read 9 November 2026). Codeleaks' list belongs to a plagiarism and licence scanner that once also returned an AI verdict; the live page keeps the heading "Source Code Detection" and eleven names, the archived FAQ of 9 December 2024 names seventeen, and the release note of 16 December 2025 withdrew the AI verdict on code (copyleaks.com/codeleaks, read 9 November 2026), so the eleven describe scanning, not AI detection. Originality.ai's seven are the languages of a benchmark sample, named because the vendor reported results on them, and the post states no supported-language list (originality.ai, read 9 November 2026). Copyleaks' product page names code as a modality and an enterprise feature without a language (copyleaks.com, read 9 November 2026), which is why the Copyleaks card and the Codeleaks card carry different fields for one company. The languages ranking counts natural languages and places Codequiry and Codeleaks at null for that reason, as each card's language note records. All read 9 November 2026; the Updated line is the date the page last changed.

How a detector treats code it was not built for, and lists, tables and references beside it, is a separate question from which languages it names: the bullet points and code question, added 9 November 2026, lists the 10 cards that state a non-prose or code limit and the 0 cards scored on code in any record.

What the language counts leave out

A named language says the vendor accepts files in it, not that the detector was tested on it. Codequiry's twelve names sit inside a claim of 65+ and the page names no sample, no accuracy figure and no false positive rate for any language (codequiry.com, read 9 November 2026). Codeleaks' eleven describe what the platform scans for plagiarism and licence matches; the AI verdict on source code was withdrawn from the submit endpoints on 16 December 2025, and the only accuracy sentence on the page is the text detector's (copyleaks.com/codeleaks, read 9 November 2026). Originality.ai's seven come with figures, 113 of 150 AI samples detected and 0 of 150 human samples flagged, but on 300 samples the vendor chose and scored (originality.ai, read 9 November 2026). The studies by content type ranking shows the code row at zero records: no outside study in the index has scored a detector on code in any language. The code use-case view sets the four cards beside the reader who needs them.