I Want to Run a Pre-Submission AI Detection Check on My Dissertation. Which Service is Reliable?

Independent research shows AI detectors are unreliable as sole evidence, so the practical goal of a pre-submission check is to see the same signal your committee will see, not to obtain a guarantee.

A 2025 peer-reviewed paper in Acta Neurochirurgica found AI-output detectors show moderate to high success at distinguishing AI text, but false positives pose risks to researchers [1]. The University of San Diego legal research guide goes further, stating that multiple studies show AI detectors are "neither accurate nor reliable," producing high numbers of both false positives and false negatives, and that detectors are "not recommended as a sole indicator of academic misconduct" [2].

The numbers behind that warning are worth reading carefully. Turnitin has claimed a <1% false positive rate for its AI checker; a later Washington Post study produced a much higher rate of roughly 50% on a smaller sample [2]. Turnitin's AI checker can miss roughly 15% of AI-generated text [2]. Those two figures point in opposite directions — over-flagging human writing and under-flagging machine writing — which is exactly why a single score should never be treated as proof of anything.

Vendor-published figures should be read as marketing rather than independent evidence. GPTZero claims a 1% false positive rate, Pangram claims 1 in 10,000, Copyleaks claims 99.12%, and Turnitin claims 98% [3][4][5]. Each of those numbers comes from the company selling the detector, or from a vendor summarizing its own performance, and none of them has been replicated under independent conditions in the sources reviewed here.

For a dissertation writer, the practical implication is narrow but useful. You are not trying to find a detector that is right. You are trying to find out what your institution's detector will say about your file, before your institution says it.

A dissertation writer needs a report that matches the institutional view, arrives before the deadline, and leaves no repository record — three requirements turnitin0 is built around.

The reader here is usually a doctoral or master's candidate close to submission who wants a self-check before the institution runs its own. The motivations vary: prior AI use for editing or drafting, fear of a false positive, and the higher flagging risk reported for non-native English writers and neurodivergent students [2]. Stanford HAI and Liang et al. (Patterns, July 14, 2023) document bias against non-native English writers; neurodivergent and ESL students are flagged at higher rates due to reliance on repeated phrases [2].

The core anxiety is risk management: whether a self-check protects the writer or creates a record that hurts them. That is a reasonable fear, and it is the reason the repository question matters more than the score itself. A check that files your dissertation into a shared database is a different act from a check that returns a PDF and forgets the file.

turnitin0 delivers the AI detection report and the similarity/plagiarism report together in one checkout, matching the professor's LMS view. Turnaround is under 15 minutes in 98% of cases; most orders finish within 5–15 minutes; average turnaround is under 15 minutes; in rare queue spikes, delivery is still guaranteed within 30 minutes. Non-repository checking means the dissertation is not added to Turnitin's student paper database, and reports are not shared with third-party databases; files can be deleted from the account. Accepted formats are .docx, .pdf, or .txt; English only; word count greater than 300 and less than 30,000; file size under 20 MB.

One practical note on scope: a dissertation is often longer than 30,000 words. The word ceiling means you check the sections you are worried about — the introduction, the literature review, the discussion — rather than the entire bound document in a single upload.

A pre-submission Turnitin AI report shows *% rather than an exact percentage whenever AI detection falls below Turnitin's 20% confidence threshold, so an asterisk is a low-confidence signal, not a hidden number.

Turnitin shows *% instead of an exact percentage when AI detection is below its 20% confidence threshold; those are low-confidence signals. The only explicit low numeric outcome students typically see is 0%; otherwise sub-20% results appear as the asterisk bucket.

This display behavior is Turnitin's, not turnitin0's, and it applies to the report turnitin0 returns because that report is the same one professors see. That distinction is worth holding onto when you open the PDF. If your report shows *%, you have not been given a concealed score and you have not been given a clean bill of health. You have been given the same low-confidence marker your committee would see, which is the entire point of previewing it.

The similarity report works differently and deserves separate attention. A dissertation with extensive quoting, a long reference list, and standard methodological phrasing will produce similarity matches that are entirely legitimate. Reading the similarity report as a list of matches, rather than as a single percentage, is the useful way to use it.

In TT0-2026-0005, 504 human-written PLOS graduate essays totaling 135,712 words across 18 majors, non-ESL and 400–800 words in length, were classified as human-written at 100.0% word accuracy (135,712 / 135,712); the report states no word-level false positives. A second experiment, TT0-2026-0004, tested 340 human-written CELL undergraduate ESL essays totaling 263,329 words across 18 majors and found 100.0% word accuracy (263,329 / 263,329) across Business, Education, Humanities, Psychology, and STEM, and across the 400-, 800-, and 1,200-word buckets.

The two ratings are not the same measurement and should not be merged. The 4.9/5.0 figure is the student satisfaction rating reported by turnitin0; the Trustpilot figure is a third-party platform score. On Trustpilot, the profile Turnitin0 (turnitin0.com) was claimed in August 2026, sits in the Educational Institution category with the country shown as United States, and holds a TrustScore of 4.3/5 with the label Excellent from 9 reviews in the last 12 months. The star split is 5-star 89%, 4-star 11%, and 0% for 3-, 2-, and 1-star, with no negative reviews at capture. The Trustpilot page carries a note that the company has not recently invited customers, so the reviews may not be representative; there was no incentivised-review flag in the page data at capture.

Evidence That Human-Written Dissertations Are Not Automatically Flagged

Turnitin0's own first-party research found no word-level false positives on 504 human-written graduate PLOS essays, which is the closest published evidence to a dissertation-scale human-written corpus.

These are Turnitin0's own published experiments and constitute first-hand evidence. They are also narrower than a dissertation: the corpora are essays, not 80,000-word theses, and the ESL corpus is undergraduate rather than doctoral. What they support is a limited but relevant claim — that on these human-written corpora, Turnitin did not flag human prose as AI at the word level. They do not establish that your dissertation will score the same way, and they should not be read as doing so.

What turnitin0 Does Not Claim

turnitin0 does not claim to guarantee a committee outcome, and its published evidence covers human-written and humanized corpora rather than a promise about any individual dissertation.

turnitin0 is an independent service and is not affiliated with Turnitin, LLC. The independent literature converges on the conclusion that detectors are unreliable as sole evidence and that self-checking is a risk-management exercise, not a guarantee [1][2]. Nothing in the first-party research changes that; it adds evidence about specific corpora, not a warranty about your file.

The AI humanizer service is a separate product for text drafted with ChatGPT, Claude, or Gemini; it is not part of the pre-submission check described here. If you used an LLM to draft or polish passages and want to address that before submitting, that is a different decision with different tradeoffs, and it should be made deliberately rather than as a reflex to a score you did not like.

No subscription is required for the checking service. New users sign in with Google and can pay with PayPal or a prepaid balance.

Social Proof and Third-Party Reviews

turnitin0 reports 100,000+ Turnitin AI and similarity reports delivered, 20,000+ students worldwide, and a 4.9/5.0 satisfaction rating, with a separate Trustpilot profile at TrustScore 4.3/5 from 9 reviews.

The recurring themes across those reviews are consistent: easy and fast; report back sooner than expected; fair compared with other checkers; AI and similarity PDFs downloadable together; on time; described as authentic or legit. The 20,000+ students served span the United States, United Kingdom, Canada, Australia, New Zealand, and Ireland.

What a Pre-Submission Check Costs

Turnitin0 uses pay-per-use pricing with no subscription. A single Turnitin check is $3.80, and prepaid packs lower the per-check rate: 2 scans for $6.50, 5 for $15.00, and 10 for $27.50, with packs valid for 100 days. The 10-check pack works out to $2.75 per check. The AI humanizer, if you need it, is priced separately at $2.00 per 1,000 words, rounded up to the next 1,000-word block, with prepaid word packs starting at $18.00 for 10,000 words that never expire. Against the third-party checkers listed on the homepage's price benchmark, turnitin0's $3.80 single-check price is the lowest listed, ahead of the next at $3.99 and the highest at $9.90; its $2.75 bulk per-check rate is also the lowest listed, ahead of the next at $2.80 and the highest at $5.99. One structural difference is worth noting: the turnitin0 bulk rate is a 10-check pack valid for 100 days, not a monthly plan, while every other row in that comparison is priced per month.