You check your assignment on Phrasly and get 7% AI. That sounds reassuring, but will Turnitin show the same thing when your lecturer checks it?
To find out, we compared Phrasly and Turnitin scores for 100 pieces of writing. The results were more useful than a simple yes or no. A very low Phrasly score matched Turnitin surprisingly well in this dataset. But as Phrasly's score increased, the two detectors disagreed a lot.
The short answer: Every one of the 22 samples scoring below 10% on Phrasly received 0% on Turnitin. At a Phrasly score below 30%, the match rate fell to 35 out of 40 (87.5%). Below 50%, it was 41 out of 50 (82%).
So a Phrasly score under 10 was a strong sign of a Turnitin 0% in our test. It is not a promise about your own submission, and Phrasly is not a substitute for a Turnitin report.
How we tested Phrasly against Turnitin
Our spreadsheet contains 100 distinct text samples, each with a Phrasly AI score and a Turnitin AI score. Both are recorded on a scale from 0 to 100. For shorter excerpts, the same passage was repeated in the Turnitin submission so that a report could be generated. We used the resulting Turnitin scores for this comparison.
We treated a Turnitin score of 0% as the main result students want to predict. We also kept Turnitin scores from 1% to 19% separate from scores of 20% or higher. That distinction matters because Turnitin's current reporting rules generally hide precise AI percentages from 1% to 19% and show *% instead. A displayed 0% is not the same thing as *%. Turnitin explains those reporting rules here.
Also Read: [STUDY] Can Phrasly AI Bypass Turnitin?
If Phrasly is under 10%, will Turnitin show 0%?
It did in all 22 cases we tested. None of the samples with a Phrasly score of 0% through 9% had a nonzero Turnitin score. Even when we included Phrasly scores of exactly 10%, the result was 26 out of 26 with Turnitin at 0%.
That is the clearest finding from the test. But 22 successes do not turn a prediction into a guarantee. If you have a Phrasly score of 8%, this data gives you a reason to be less concerned about a different Turnitin result. It cannot tell you what Turnitin will return for every new essay.
Also Read: [STUDY] Can Phrasly AI Humanizer Outsmart Sapling? A Data-Driven Test
What about Phrasly scores below 20%, 30% and 50%?
Here is the full comparison using cumulative cutoffs. "Below 30" means 0% through 29%, not 30% itself.
| Phrasly AI score | Samples | Turnitin 0% | Turnitin 1%-19% | Turnitin 20%+ | Turnitin 0% rate |
|---|---|---|---|---|---|
| Below 10% | 22 | 22 | 0 | 0 | 100% |
| Below 20% | 37 | 32 | 1 | 4 | 86.5% |
| Below 30% | 40 | 35 | 1 | 4 | 87.5% |
| Below 50% | 50 | 41 | 1 | 8 | 82.0% |
| Below 70% | 66 | 54 | 1 | 11 | 81.8% |
The groups overlap. For example, the 22 samples below 10% are also part of the 40 samples below 30%. The final column is the share of each group that received a recorded Turnitin score of exactly 0%.
Why is the below-30% result slightly better than below 20%? Three more samples entered the group, and all three received 0% on Turnitin. With a dataset this size, the percentages do not have to move smoothly.
The important change happens once you move past the 10% range. Among the 15 texts scoring from 10% to 19% on Phrasly, 10 received Turnitin 0%, one received 19%, and four received 20% or higher. For example, one passage scored 14% on Phrasly but 100% on Turnitin. Another scored 19% on Phrasly and 94% on Turnitin.
That is why a Phrasly score of 15% or 25% should not be treated as a reliable forecast of Turnitin 0%, even if many examples in those ranges did end up at zero.
Do Phrasly and Turnitin usually give similar percentages?
No. Across all 100 samples, Phrasly's average AI score was 49.8%, compared with 27.9% for Turnitin. Phrasly gave the higher score in 69 cases. Turnitin gave the higher score in 15 cases, and the two were identical in 16.
The scores did have a positive relationship, but it was far from a close match. Their Pearson correlation was about 0.47. In plain English, a higher Phrasly score tended to come with a higher Turnitin score, but there were plenty of exceptions.
One real example from the dataset was a passage about an infection prevention program in European hospitals. Phrasly assigned it 59%, while Turnitin assigned it 100%. Here is the Phrasly check for that passage.
What if Phrasly says 100% AI?
This was one of the biggest surprises. Phrasly marked 29 of the 100 samples as 100% AI, but Turnitin did not give all of those samples high scores.
Here is what Turnitin reported for those 29 texts: 10 scored 0%, three scored 18% or 19%, three scored between 20% and 99%, and 13 scored 100%.
In other words, 10 out of 29, or 34.5%, got Turnitin 0% even though Phrasly was completely certain on its own scale. So a Phrasly 100% is not proof that Turnitin will return 100%, or even that Turnitin will show a nonzero percentage.
The mismatch goes both ways. Turnitin gave 21 samples a score of 100%. Four of those scored below 50% on Phrasly, including one that Phrasly scored at just 14%.
Also Read: [STUDY] Is Pangram AI Detector Similar to Turnitin?
Can students use Phrasly as a Turnitin score checker?
It can be a rough indicator, especially at the very bottom of the Phrasly scale, but it is not a Turnitin score preview. This test suggests three practical ways to read a Phrasly result:
- Below 10%: This was the most dependable range in our test. All 22 samples received Turnitin 0%. Still, do not assume the same result is guaranteed for a different submission.
- 10% to 49%: Treat the score with caution. We saw both Turnitin 0% results and much higher scores, including Turnitin 100%.
- 50% to 100%: A high Phrasly score does not tell you exactly what Turnitin will say. Of the 50 samples scoring 50% or more on Phrasly, 24 received Turnitin 0%.
Also remember that the Turnitin AI score is separate from its similarity score. Neither tool can establish who wrote a passage just from a percentage. Turnitin itself advises instructors not to use AI detection alone as the basis for an adverse decision.
Our test measures agreement between two detectors, not whether either detector correctly identifies the true author. The sample contains 100 paired results, not a guarantee for every subject, writing style or future software version. Turnitin also updates its detection model, so a test like this is a snapshot.
Frequently asked questions
If Phrasly shows 0% AI, will Turnitin also show 0%?
It did for all three samples in our dataset that received exactly 0% on Phrasly. Three cases are too few to make a firm prediction for everyone.
What if Phrasly is exactly 10%?
Four samples had a Phrasly score of exactly 10%, and all four received Turnitin 0%. Counting them with the scores below 10 gives 26 out of 26 at Turnitin 0%.
Is a Phrasly score below 30% safe for Turnitin?
Not reliably. Of the 40 samples below 30%, five received a nonzero Turnitin score. Four of those were at least 20%, including one Turnitin 100% result.
Can Phrasly 100% mean Turnitin 0%?
Yes. That happened in 10 of the 29 samples Phrasly rated at 100%.
Does Turnitin 0% mean the work is definitely human-written?
No. A Turnitin 0% means its model did not identify qualifying text as likely AI-generated in that report. It is not a certificate of authorship. Similarly, a high score alone is not proof of misconduct.
Bottom Line
If your Phrasly score is below 10%, our results are encouraging: 22 out of 22 samples also came back as Turnitin 0%. Once you move into the teens and above, the match becomes less predictable. By the time you include everything below 50%, nine of the 50 samples have a nonzero Turnitin score.
Use Phrasly as one imperfect signal, not as a stand-in for Turnitin and certainly not as a test of whether your writing is genuinely your own. If you wrote the work yourself, keeping drafts, notes and revision history is more meaningful than chasing a particular detector percentage.