Your international recruitment target went up. Your writing support establishment did not.
That is a familiar budget story, and it would be an ordinary capacity problem except for one thing: the gap it opens does not land evenly. It concentrates on a single cohort, and it produces two different institutional exposures at the same time — one about fairness and one about integrity. Most institutions manage them in separate committees and never notice they are the same problem.
The two exposures, and why they converge
The first is an equity exposure. A candidate writing a thesis in a second language needs more editorial support than one writing in a first language. That is not a comment on their research, their intelligence or their preparation; it is a straightforward fact about producing 80,000 words of disciplinary prose in an acquired language. If the institution does not supply that support, it is bought privately — and the University of Toronto’s School of Graduate Studies states the consequence in its own guidance, that engaging a copy editor for a graduate thesis “can be a significant financial burden.”
Where provision runs out, the polish of a thesis starts to track the ability to pay. Two candidates with equally good research submit differently finished documents, and the difference is a bank balance.
The second is an integrity exposure, and it runs the other way. AI detectors are documented as being more likely to flag text written by non-native English speakers — one of the reasons Vanderbilt University cited when it disabled Turnitin’s AI detection tool, as set out in whether AI detection is reliable enough to base a case on. An aggregate false-positive rate conceals a materially higher rate for this cohort.
So the same students are simultaneously the most likely to need paid help, the least likely to be able to afford it, and the most likely to be wrongly accused by the system that screens the result. An institution with an equality duty cannot treat any of those three as an acceptable error term.
The design error underneath it
Most institutions do have international student support. It is usually pre-sessional or in-sessional English language provision, it is often good, and it is answering a different question.
Language support addresses grammar, idiom, register and fluency. It is front-loaded, delivered before or early in a programme, and assessed against language proficiency.
Doctoral writing support addresses how an argument is built across a chapter, what a literature review is for, how a discipline signals a claim, and how eighty thousand words hold together. It is needed in year two and year three, not in induction week.
A candidate can pass every language requirement, speak and write fluently, and still need the second — because the second is disciplinary rather than linguistic. When an institution points a struggling third-year candidate at the language centre, it is offering a service that does not address the problem, and the candidate concludes, correctly, that there is nothing internal for them.
That conclusion is the moment the private purchase happens. What they buy next is described in institutional writing support versus external editing services, and the shape of the market is not uniform: some of it is qualified academic editing that would decline to touch an argument, and some of it will rewrite a chapter and call it proofreading. The student choosing between them is the least equipped person in that transaction to tell which is which.
Why the supervisor cannot absorb it
The standard institutional answer is that the supervisor provides writing development, and in principle that is right — supervisors are usually named as the primary resource.
In practice, supervisors are the constraint. Where a candidate needs sustained language-level and structural feedback, a supervisor either spends review time on prose instead of research — the workload problem described in what your supervisors are actually reviewing — or stops commenting on it, which leaves the candidate to discover the gap at examination. Neither is a support model, and both concentrate on the same cohort.
What Tesify for Institutions does about it
The narrow claim first: this is provision, not screening, and it addresses the gap between language support and supervision rather than replacing either.
- It is available at the point of writing, at three in the morning, in year three. That matters more for this cohort than for any other, because the need is continuous rather than concentrated in induction and because appointment-based provision is exactly what a waiting list rations first.
- It works on structure and consistency, not only sentences. Chapter architecture and referencing held stable across a document too long to hold in your head is the disciplinary layer that language provision does not cover.
- It costs the student nothing. An institutional licence removes the ability-to-pay variable from the outcome, which is the equity exposure resolved directly rather than mitigated.
- It leaves a record of how the document was produced. For a cohort disproportionately exposed to false flags, an inspectable trail is protective. A candidate who can show the drafting is in a far stronger position than one holding only a finished PDF and a denial.
- It is a disclosed, institutional route. Help that goes through your platform is help you can see, scope and describe in a policy — unlike the same help bought privately.
What we do not claim: that a platform substitutes for teaching, for supervision, or for a properly resourced writing centre. It scales the layer that cannot be staffed, and it makes the rest visible.
Four questions worth answering before the next intake
- What is the international share of your postgraduate research population, and how has it moved in five years? Compare it with the change in your writing support establishment over the same period. The two numbers are rarely on the same page and they should be.
- Does your provision distinguish language support from disciplinary writing support? If a third-year candidate is referred to the language centre, the answer is no.
- What is the waiting time for your writing support, and does it differ by cohort? If you have never disaggregated it, that is the first measurement to take.
- If you screen with a detector, do you monitor flag rates by cohort? An aggregate rate hides a differential one, and the differential is the equality question.
Question four is the uncomfortable one and the cheapest to answer. It requires no new system, only a breakdown of data you already hold.
Start with a departmental pilot, in the right department
Pick a programme with a high international intake and a head who will engage, agree the success criteria before it starts, and include a comparison group so the result means something. It commits coordination time rather than budget, and it clears approval routes that a purchase order does not.
Instrument it properly or it will produce an impression rather than a finding — the design is in how to run a departmental pilot of an AI writing tool, and the evaluation method, including why attender-versus-non-attender comparisons mislead, is in how to measure whether a writing support intervention actually worked.
If your international intake has grown faster than your writing provision, request an institutional evaluation and we will scope a free departmental pilot against a programme you name, with the cohort breakdown agreed up front.
Frequently asked questions
Is this not just saying international students write badly?
No. It is saying that producing a thesis in a second language takes more editorial effort, which is a fact about the task rather than about the researcher. The research quality is not the variable in question; the finishing support is.
Why is the integrity risk higher for this cohort?
Because detectors are documented as more likely to flag text written by non-native English speakers, and because unmet support needs are more likely to be met by an undisclosed private purchase. Both effects concentrate on the same students.
Does our existing English language provision cover this?
Usually not. Language provision addresses fluency and is front-loaded; doctoral writing support addresses disciplinary argument and is needed in years two and three.
What does an institutional licence cost?
It depends on population, scope and term, and no vendor in this category publishes a static institutional price. We quote against written assumptions you set so the figure is comparable with any other quotation. What to require is in what an institutional licence should actually buy you.
Where is our data processed?
Set per institution in the data processing agreement, with a sub-processor list naming the location of each. The reason a webpage claim is not sufficient is covered in where your students’ text is actually processed.
How do you handle GDPR and FERPA?
Through the data processing agreement, with a lawful basis, a minimised field set, a retention period and a sub-processor list, reviewed through your own process rather than ours.
How much integration effort is required?
SSO and standards-based LMS integration, without a separate implementation fee. A pilot can run on a narrower configuration.
Does providing a writing tool undermine academic integrity?
Provided, disclosed and visible support is the opposite of the risk. The integrity exposure comes from help that is bought privately, scoped by nobody and recorded nowhere — which is what happens by default when provision is absent.
Should support be targeted at international students specifically?
Be careful. Targeting reaches the need efficiently but can label a cohort, and domestic students need disciplinary writing support too. Most institutions need a universal baseline with additional capacity where need concentrates.
How do we know whether the gap exists at our institution?
Disaggregate three things you already hold: writing support waiting times, integrity case rates and detector flag rates, each by cohort. If none is currently broken down, that is the finding.
Will this replace our detection contract?
No. Screening examines work after submission; support changes how it is produced. They answer different questions and a displacement business case will not hold.
What does the pilot commit us to?
Coordination time and pre-agreed success criteria. No budget, no procurement cycle, and it runs alongside your existing contracts.