A sentence-level pre-Turnitin review for graduate seminars, dissertation chapters, and qualifying exams, plus a private pre-submission scan for your own journal papers, grants, and conference drafts. Calibrated for the formal register, methods-section phrasing, and theoretical framings that detectors trained on undergraduate writing tend to over-flag. Free to try. No card. Your first scan in about six seconds.
Undergraduate seminars, master's coursework, and doctoral supervision each carry a different shape of AI review in 2026. TextSight is tuned for all three contexts.
The pressure here is volume. A 30 to 80 student section drops papers in waves, and faculty want to know which submissions to read closely before grading. The TextSight workflow sorts the class by Authenticity Score, surfaces the bottom decile, and gives sentence-level highlights on each flagged paper, so the office-hours conversation is anchored on specific passages rather than a vague sense that something feels off.
Higher stakes per submission, longer documents, and a stronger student incentive to use AI substantively. Pro accepts longer passages per scan, which covers most capstone chapters in two or three passes. The sentence-level evidence travels across passes, so a 40-page methods chapter can be reviewed in pieces without losing the picture of the whole.
The supervisor relationship is built on trust, and a flagged scan is a delicate signal to raise. The right pattern is to scan chapters as drafts arrive, log results privately, and bring up concrete paragraphs in the supervision meeting only when the picture is clear. The 90-day history keeps the trail intact across a semester of chapter cycles without manual archiving, and the saved scan PDFs anchor the file if a formal review becomes necessary later.
Turnitin and iThenticate remain the institutional record. TextSight runs faster and earlier, with different signals, so faculty can read evidence rather than a single percentage.
Most universities run Turnitin's AI check or iThenticate as the official screen at submission. Those tools produce a percentage and an institutional record, which is the right shape of evidence at the policy layer. They are slow during peak grading weeks, the output is shallow, and the percentage does not tell a faculty member which paragraphs to look at.
TextSight runs the scan in seconds rather than minutes, surfaces sentence-level highlights with per-sentence confidence, and exposes the signals the classifier weighted. The percentage is similar in spirit but the working unit is the sentence, which is the unit a faculty member can act on. When the institutional Turnitin report lands later in the week, the two views together produce a much sharper picture than either does alone.
Each sentence is scored independently and colour-coded by confidence band, so the paragraphs that cluster red are the ones to read closely with the candidate, while the paragraphs that come back clean give a clear baseline. The classifier also exposes the underlying signals: perplexity reads how predictable each token is given the surrounding context, burstiness reads sentence-length variance across the document, and sentence-variance signals catch the narrow band AI-generated prose tends to land in. Each signal is shown individually so a faculty member can see why a passage flagged, not only that it did.
Paste a chapter section by section, and the result panel keeps each pass in the same workspace. Section-level scores plus sentence highlights rebuild the picture of a 40-page document without losing precision, and the saved scans link together in the 90-day history for later reference.
The classifiers under Turnitin and TextSight were trained on different corpora and use different signal weights. A passage that lands clean on one and flagged on the other is itself useful information, which usually means the writing sits in a borderline zone of academic register and benefits from a closer read. Faculty who run both as part of the standard workflow report a higher confidence in their grading calls and a lower false-positive escalation rate.
Use the free tier today, no email needed. Paid tiers billed in USD; institutional invoicing available on Business. Full details on the pricing page.
Billed $89.88/year — Save $30
Billed $179.88/year — Save $60
Billed $359.88/year — Save $120
Yearly billing saves 25%. View full pricing →
Most US R1, UK Russell Group, and EU research departments wrote formal AI policies between 2024 and 2026. The standard shape of the policy is documented evidence, a faculty conversation, and human judgement before any referral.
Run the submission through TextSight after grading begins but before any conversation with the student. Save the scan to the 90-day history. If the score is clean and no paragraphs cluster red, the document is ready for normal grading. If the picture is mixed, treat it as a signal to read the work more carefully rather than as evidence of intent.
The sentence-level highlights point to concrete paragraphs. The office-hours conversation is "walk me through your argument in section three" rather than "this feels AI-generated." The candidate has a fair chance to address the evidence and to explain their drafting process. The conversation, not the scan, is the primary evidence of the integrity call.
If the conversation does not resolve the question, the saved scan PDF anchors the file for a formal referral. Most departments require evidence beyond a single classifier score, so the scan sits alongside the conversation notes, earlier drafts, and any institutional Turnitin or iThenticate output. Faculty judgement outweighs the classifier in every reasonable process.
Departments still drafting policy in 2026 often pull from faculty scan logs as precedent. Saving scans privately during the policy gap is itself useful institutional contribution.
No native LMS plugin yet. The current workflow is copy-paste or PDF upload into the TextSight web app, which most faculty find faster than an in-LMS plugin for typical seminar volumes.
The standard workflow is to download the submission from your LMS as a PDF or paste the text directly into TextSight. Result panels save into the 90-day history with the original input, so a scan run on a Canvas submission stays linked to that submission in your private history. Native Canvas and Moodle plugins are on the 2026 roadmap. Blackboard and Brightspace are slated to follow once the Canvas integration is shipped and stable.
Business tier includes REST API access, which institutions running scripted grading pipelines can use to scan submissions in batch. The API returns the same sentence-level evidence the web app shows, so a department writing custom Canvas LTI integration can wire TextSight into the grading queue without waiting for the native plugin.
Turnitin's AI check and iThenticate stay as the institutional submission-layer record. TextSight runs alongside as the working tool during grading, which is where the sentence-level evidence pays off. The two are complementary, not duplicate, and most faculty workflows use them in sequence.
Faculty sit in two chairs in 2026. TextSight is built so the same tool serves both, with the workflows kept separate inside one account.
Scan dissertation chapters as drafts arrive. Log the result privately. Surface concrete paragraphs in the supervision meeting only when the picture is clear, and treat the conversation as the primary evidence of any concern. The 90-day history keeps the trail intact across a semester of chapter cycles without manual archiving, and the saved PDFs link to the file if a formal review board ever needs them later.
Nature began AI screening in 2024. Science, IEEE, ACS, and most major medical journals followed by 2025. Editorial-desk screening produces a number and a section-level breakdown, often before the paper reaches a human editor, and a flag at the screening stage can delay decisions by weeks. Run your draft through TextSight before submission, see which paragraphs the screener will likely flag, and either rewrite them in your own register or use the in-product AI rewriter to reset the language while keeping your content intact.
Some faculty use TextSight to spot-check sections of their own published work, usually before a press interview or a public talk where the underlying methods will be scrutinized. The 90-day history keeps the audit accessible, and the saved PDFs are useful evidence if a post-publication AI question lands in your inbox.
NIH and NSF do not screen for AI today, but several program officers have signalled that screening is coming. ACM-track and several IEEE conferences already run submissions through detectors during desk review. A pre-submission scan adds five minutes to the workflow and removes a class of avoidable problems.
Student work is sensitive. TextSight is built so faculty can use it without rerouting student text into a training pipeline.
Submitted text is never used to train the TextSight classifier. The classifier was trained on internally licensed academic corpora plus public-domain writing, with student work explicitly excluded. Pasting a graduate seminar paper into TextSight does not contribute that paper to a future classifier release.
Every scan is logged in your account and viewable only by you. Pro keeps 90 days of history, Business keeps history indefinitely with optional deletion on request, and the audit log on Business is suitable for academic integrity review boards that want a consistent record across multiple graders.
Standard DPA available on Business and Enterprise. Data residency on request for institutions with stricter requirements. Faculty should still confirm their own department or IRB position before routing protected human-subjects research data through any third-party detector, since the policy varies by institution and by research category.
Same engine, different workflow for short-essay grading at high volume with parental involvement.
K-12 workflow →Department-level licensing, shared scan history, and audit-log workflows for review boards.
Institution licensing →Clearing journal AI screening on language-polished or methods-scaffolded drafts.
Read the guide →Full tier breakdown for Free, Starter, Pro, and Business. Annual billing saves 25%.
See pricing →Free to try. No card. Sentence-level highlights, 90-day private history on Pro, and a built-in AI rewriter for your own pre-submission drafts.