Comparisons

Academi.cx vs ZeroGPT 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

ZeroGPT offers a free AI checker and advertises 98.4% detection accuracy with a false positive rate below 1%. Those are confident claims, but students checking an assignment still need to know whether its results resemble what Turnitin will report. We put ZeroGPT and Academi.cx through a direct comparison with Turnitin to find out which comes closer.

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, ZeroGPT, and Turnitin. We compared how closely Academi.cx and ZeroGPT matched Turnitin’s overall AI scores and sentence-level highlights.

The results from each detector

AI scores for five documents. Document 1: Academi.cx 46%, ZeroGPT 40%, Turnitin 55%; predicted range 37 to 42%. Document 2: 48%, 42%, and 74%; range 42 to 52%. Document 3: 59%, 53%, and 56%; range 42 to 63%. Document 4: 70%, 57.2%, and 65%; range 62 to 74%. Document 5: 74%, 55.2%, and 71%; range 68 to 78%. Blue circles show Academi.cx, gray squares show ZeroGPT, navy diamonds show Turnitin, and capped blue lines show predicted Turnitin ranges.
  • Document 1: Academi.cx detected 46%, predicting 37 to 42%. Turnitin's actual score was 55%, above that range. ZeroGPT detected 40%, 15 points below Turnitin.
  • Document 2: Academi.cx detected 48%, predicting 42 to 52%. Turnitin returned 74%, well above the range. ZeroGPT detected 42%, missing Turnitin's score by 32 points.
  • Document 3: Academi.cx detected 59%, predicting 42 to 63%. Turnitin's actual score was 56%, inside the range. ZeroGPT detected 53%, putting both detectors 3 points away from Turnitin.
  • Document 4: Academi.cx detected 70%, predicting 62 to 74%. Turnitin returned 65%, inside the range. ZeroGPT detected 57.2%, 7.8 points below Turnitin.
  • Document 5: Academi.cx detected 74%, predicting 68 to 78%. Turnitin's actual score was 71%, inside the range. ZeroGPT detected 55.2%, 15.8 points below Turnitin.

Where Turnitin scored higher

Academi.cx's predictions came in below Turnitin on the first two essays, particularly the shorter second document, where a handful of sentences made a bigger difference to the final percentage. Some of the text Turnitin flagged still appeared in our report, but Academi.cx gave those sentences lower AI likelihoods, so they did not count toward the overall score. That left us with a lower percentage even though the report still showed much of the text Turnitin had flagged.

Average deviation from Turnitin across all five documents: Academi.cx 9.2 percentage points; ZeroGPT 14.7 percentage points. Values are mean absolute differences, rounded to one decimal. Bars share a zero baseline and are proportional.

How the sentence highlights compared

ZeroGPT removed headings, bold text, and indentation from its report, which made the essays harder to follow alongside Turnitin. When we compared the marked passages, ZeroGPT left out parts of paragraphs Turnitin flagged and marked other text Turnitin left alone, while Academi.cx generally followed Turnitin more closely. ZeroGPT was closer on a paragraph in document four and a short section of document five where Academi.cx marked too much, although both detectors also flagged some text Turnitin did not.

The final scorecard

Academi.cx took all five rounds on the overall scores and sentence highlights, with its reports showing more of the text Turnitin flagged across the test. The predictions on the first two essays were too low, but the sentence-level detail still gave a clearer idea of which parts of the writing Turnitin was likely to highlight. ZeroGPT did better on a few individual passages, although it left out more of Turnitin's flagged text overall and was less consistent from one essay to the next.

Related reading