Here’s Why Universities Are Ending Turnitin Contracts

A university lecture hall with laptops, symbolizing the debate over Turnitin's plagiarism and AI-detection software in higher education

For years, Turnitin was the gatekeeper of academic honesty. Submit an essay anywhere, and its software would scan it against billions of web pages and student papers, spitting out a similarity score. Now, universities are walking away. In 2026, several British institutions — Newcastle, Southampton, and Westminster among them — announced they will end or let lapse their Turnitin contracts, driven by data-privacy fears, unreliable AI detection, and a core product that generative AI has left behind.

Key Takeaways

  • UK universities including Newcastle, Southampton, and Westminster are dropping Turnitin after proposed licence changes raised fears student work could be used for AI purposes.
  • Westminster called Turnitin’s AI detector ‘probabilistic rather than conclusive’ with a real risk of false positives against genuine student work.
  • Experts say the deeper driver is money and obsolescence: AI has made classic plagiarism checking largely redundant.

The trigger: a licence change about student data

The immediate spark was contractual. Turnitin proposed updating its End-User Licence Agreement (EULA), with new terms on processing student submissions with AI tools. Universities read that as a potential green light for student and staff work to be used for AI purposes — and they pushed back hard.

Newcastle University’s education committee decided on a “phased withdrawal” of the service, ending it after the 2026-27 academic year, citing those concerns. The University of Southampton reached the same conclusion earlier in 2026, saying it would not renew once its current deal lapses.

Turnitin has denied the worst fears, saying student work is not used to train any large language model — only a classifier that predicts whether text looks AI-written. After sector-wide backlash through the university IT body UCISA, Turnitin postponed the EULA changes to September 2027 and agreed to redraft them in partnership with UCISA.

But the damage was done. The University of Cambridge refused to sign the new terms, saying it would “not accept any changes to the EULA that impact on data privacy or intellectual property,” and is operating under a modified arrangement while exploring other software. Aberystwyth — contracted until July 2027 — is now telling students they can submit and be marked through Blackboard instead if they are uncomfortable with Turnitin. Aberdeen is reviewing its position.

The deeper problem: the tool no longer fits the threat

Data privacy was the trigger, but experts say the real driver runs deeper: what is Turnitin actually protecting against anymore?

Westminster gave the bluntest answer when it announced it would retire Turnitin in July 2027. Its core plagiarism checker, the university said, was “largely ineffective against the most significant current integrity risk” — students using generative AI to produce coursework. A similarity checker looks for copied text; an essay written from scratch by ChatGPT copies nothing, and sails through.

Turnitin’s newer AI-detection feature hasn’t filled the gap. Westminster described it as “probabilistic rather than conclusive” and warned of a “recognised risk of false positives” against genuine student work — and detectors have indeed flagged human-written text as AI-generated, in some cases as 100% AI-written when it wasn’t.

A global revolt against AI detectors

The scepticism is not limited to the UK. In the United States, a growing list of institutions — including Yale, Johns Hopkins, Indiana, Northwestern, Georgetown, and New York University — have enacted policies that ban or discourage faculty from relying on AI-detection tools as the sole evidence of cheating. At least a dozen universities have disabled Turnitin’s AI detection feature entirely, after its false-positive rate turned out to be higher than the company originally suggested.

Indiana University’s Kelley School of Business put it directly in its faculty playbook: AI detectors are “highly unreliable,” and professors should instead design assignments that encourage process, reasoning, and authentic engagement. There is also a cat-and-mouse problem: “humanizer” tools defeat detectors almost as fast as they improve, turning integrity into a technical arms race instead of a learning event.

The money question

Some academics argue finances are the deeper driver. Thomas Lancaster of Imperial College London has suggested that universities are using AI concerns to justify walking away from contracts when saving money is the real motivator.

The costs are not trivial. Freedom-of-information releases show the University of Southampton paid £87,621 for Turnitin between September 2024 and September 2025, while the University of East Anglia paid £50,892 over a similar period. Turnitin’s prices aren’t published and vary by institution and features — and the company now bundles capabilities many universities feel they neither need nor want to pay for.

Others push back. Phil Newton of Swansea University has said the licence fee is not a major burden in the grand scheme of a university budget — though with UK universities under strain, any savings are welcome. But Newton made a sharper point: most UK universities only use Turnitin’s originality-checking function, which cannot detect AI-generated text at all. Paying for a tool that misses the dominant form of academic misconduct is a hard sell, whatever the price.

What comes instead

Universities are not abandoning academic integrity — they are unbundling it. Platforms like Blackboard and Canvas already handle submission, grading, and feedback natively, so institutions can drop Turnitin’s bundled workflow and plug in lighter originality-check alternatives where they still want them.

The bigger shift is pedagogical: oral exams, in-class writing, and process-based assignments where students show drafts and reasoning. The detector era produced false accusations without stopping cheating; redesigning the assessment removes the incentive to cheat in the first place.

What this means for Pakistan

This matters for Pakistani readers because Turnitin is deeply embedded in the country’s higher education system. HEC has issued detailed SOPs governing Turnitin accounts across institutions, uses Turnitin itself for preliminary checks in plagiarism cases, and its anti-plagiarism policy treats the similarity index — allowed up to 19% — as a standard gate for theses and research manuscripts. For PhD scholars, clearing the Turnitin check is a non-negotiable step before a thesis reaches external examiners.

No Pakistani university has announced plans to drop Turnitin, and HEC’s framework still assumes it. But the global concerns apply here just as strongly: if student theses train AI systems, data-privacy questions would hit Pakistani research too. And AI-detector reliability is arguably worse for Pakistan’s students — studies show detectors are biased against non-native English writers, a category that includes most Pakistani students. A false accusation of AI cheating can derail a degree.

The lesson Pakistani institutions should take from the UK exodus: separate the pieces — use similarity tools for what they do well, never treat an AI-detection score as proof of misconduct, and watch the EULA terms of any vendor holding thousands of Pakistani theses on its servers. The universities ending their Turnitin contracts are not giving up on integrity — they are admitting that a similarity score was never the same thing.