iThenticate Review (2026): What It Checks and What It Cannot Decide

Turnitin explains that the overall similarity percentage is the proportion of text matching other sources in its database. Matching text can include quoted material and references. Its report interface can group matches by whether it recognizes citations or quotation marks, but the company also warns that these recognition features do not get every case right.

That makes the report a review queue, not a verdict. Open the largest or most consequential matches first, compare source and manuscript context, then decide whether the passage needs quotation, attribution, rewriting, disclosure of prior dissemination, or no change. Do not rewrite merely to drive down a percentage if the underlying source use is already accurate.

Sources: iThenticate Similarity Report guide, iThenticate and plagiarism guidance, checked July 15, 2026.

What it does well: It gives a researcher a documented view of matching text and sources, including tools for examining overlap and applying report filters. That is valuable when the manuscript contains prior work, quotations, densely cited background, or multiple author contributions that need a focused overlap review.

Where it falls short: It compares text. It does not assess experimental design, numerical analysis, image integrity, factual accuracy, citation entailment, author contribution, or whether a manuscript makes a defensible scientific contribution. Its database and account settings also define what the report can surface, so a lack of a match is not proof that no scholarly issue exists.

Best alternative route: Use a citation or manuscript review workflow when the concern is claim support, methods, reporting completeness, or journal fit rather than wording overlap. Use the target journal, institution, or funder policy when the concern is a formal integrity decision.

A manuscript with substantial prior work from the same group. Compare methods, introductions, figure captions, and discussion language against earlier papers. The goal is not a low score. It is an accurate account of reused protocol text, prior findings, and what is genuinely new in the present manuscript.

A review or evidence synthesis with dense quotation and citation. Use the report to locate overlap that may be expected, then check that direct wording is clearly marked and that synthesis is not merely source-by-source paraphrase.

A multi-author paper with uneven drafting history. A report can give the corresponding author a concrete list of passages to review before submission. It cannot establish who wrote a passage or whether a scientific claim is defensible.

In our pre-submission review work, we see authors treat a low similarity score as a general readiness signal. What actually happens is that the report has answered a narrow text-overlap question while the manuscript still leaves a scientific or editorial problem untouched. These are specific named failure patterns, not claims about iThenticate's private decision logic.

We do not treat an iThenticate report as a Manusights scientific review, and we do not infer a journal outcome from it. We observe that the useful handoff is concrete: use the report to identify a passage, then inspect the manuscript component it touches. A methods match may require a prior-work disclosure; a citation match may require checking whether the source supports the surrounding claim; a repeated abstract phrase may require a clearer explanation of what is new in the current paper. That process preserves the source context rather than optimizing a dashboard number.

For iThenticate users, the final audit should include the text that will actually be submitted, the references and quotations as formatted for the target journal, and any supplements or cover letter that state the relationship to prior work. A similarity report cannot reconcile contradictions between those documents. The author team must make the publication history, contribution, data availability, and disclosure record consistent across the package.

A low similarity score used as evidence that citations are correct. A paragraph can contain little copied text and still cite a source that does not support its claim, omit a necessary citation, or overstate the evidence. Check claim-to-source support separately.

A high score reduced by mechanical rewriting. Synonym swaps can lower textual overlap while preserving an inaccurate attribution, a misleading paraphrase, or a recycled argument. Read the source and manuscript together; fix the scholarly problem, not the color indicator.

A methods match hidden without a prior-work explanation. Reused protocol language can be legitimate, but an editor or reader may still need to know how the present work differs from a preprint, registered protocol, thesis, or earlier article. Make that relationship explicit where the target policy requires it.

An AI-writing indicator treated as a misconduct conclusion. Turnitin says its AI-writing detection can misidentify text and should not be the sole basis for adverse action. It is separate from similarity reporting and requires human judgment under the applicable policy.

Turnitin describes AI-writing reporting as separate from the Similarity Report and available only where the account has the capability enabled. Its current guidance says the model may misidentify human-written, AI-generated, or AI-paraphrased text and should not be the sole basis for an adverse action. File, language, length, and account conditions also affect whether a report is produced.

That boundary matters for researchers. A prose classifier cannot establish who generated a result, whether an analysis is valid, whether an author followed a journal's disclosure policy, or whether a manuscript's substantive claims are original. Read the current journal policy and preserve an accurate record of tool use and author responsibility.

We reviewed current public iThenticate product and report-interpretation documentation, plus independent university guidance describing researcher use before publication. We did not test a private account, upload a manuscript, buy a subscription, or benchmark iThenticate against another detector. This review therefore assesses documented workflow and limits, not private report quality, coverage, or pricing.