Comparisons

Academi.cx vs Originality AI Accuracy Test

We compared each detector’s overall AI scores and sentence-level highlights with Turnitin’s results across five essays.

Brayden · 2 min read

Originality.ai places a particular emphasis on detecting AI writing that has been paraphrased or edited. Built for writers, publishers, and educators, it claims leading detection accuracy backed by third-party studies. We wanted to see how its approach lines up with Turnitin, so this comparison puts Originality.ai alongside Academi.cx to find out which more closely matches Turnitin’s results.

Watch the full comparison on YouTube

How the test worked

Five essays were sourced from IvyPanda, a free essay database with no affiliation to Academi.cx. Selected paragraphs in each essay were rewritten using ChatGPT to introduce AI-generated content. The same version of each document was then submitted to Academi.cx, Originality AI, and Turnitin. We compared how closely Academi.cx and Originality AI matched Turnitin’s overall AI scores and sentence-level highlights.

The results from each detector

Results for five documents with blue Academi.cx scores and predicted ranges, Originality AI scores, and navy Turnitin scores. Document 1: Academi.cx 20%, range 11–25%; Turnitin 31%; Originality AI 98%; Document 2: Academi.cx 38%, range 32–42%; Turnitin 35%; Originality AI 69%; Document 3: Academi.cx 28%, range 20–33%; Turnitin 26%; Originality AI 61%; Document 4: Academi.cx 0%, range 0–19%; Turnitin *%, exact result hidden; Originality AI 32%; Document 5: Academi.cx 73%, range 64–78%; Turnitin 48%; Originality AI 100%. *%: Turnitin hides results above 0% and below 20%. Originality AI reports AI probability.
  • Document 1: Academi.cx detected 20%, predicting 11 to 25%. Turnitin’s score was 31%, above the range. Originality AI returned a 98% AI probability.
  • Document 2: Academi.cx detected 38%, predicting 32 to 42%. Turnitin’s score was 35%, inside the range. Originality AI returned a 69% AI probability.
  • Document 3: Academi.cx detected 28%, predicting 20 to 33%. Turnitin’s score was 26%, inside the range. Originality AI returned a 61% AI probability.
  • Document 4: Academi.cx detected 0%, predicting 0 to 19%. Turnitin returned *%, indicating a nonzero score below 20% with no highlights. Originality returned 68% original (32% AI probability).
  • Document 5: Academi.cx detected 73%, predicting 64 to 78%. Turnitin’s score was 48%, below the range. Originality AI returned a 100% AI probability.

Where Originality AI highlighted more text

Originality AI often highlighted whole paragraphs where Turnitin had marked only a few sentences or left the text unflagged. On the second essay, it also left some of the opening paragraph unmarked where both Academi.cx and Turnitin had flagged AI, before highlighting additional passages further down. Academi.cx followed Turnitin more closely on those sections and on several passages in the third essay, although we also marked some extra sentences.

Average deviation from Turnitin: Academi.cx 10.3 percentage points; Originality AI 47.0 percentage points. Paired rounded bars share a common zero and scale. Four visible Turnitin results; score agreement, not Originality AI detection error.

Two predictions fell outside the range

Academi.cx's first prediction came in too low, with some opening sentences Turnitin flagged not clearly identified in our report, while the last essay went in the other direction after we highlighted an additional paragraph and several extra sentences. On the fourth essay, Turnitin returned a hidden result within our predicted range, but without its highlights there was no way to compare the passages it flagged with the text shown in our report.

The final scorecard

Academi.cx's scores were closer to Turnitin on every essay where an exact result was available, and its sentence highlights generally followed more of the same passages. Originality AI detected AI content throughout the test, but often highlighted larger sections than Turnitin did, while our own predictions were still too low on the first essay and too high on the last.

Related reading