When researchers and students search for “AI writing tools approved by universities,” they’re usually looking for something more concrete than a policy framework: a list of specific tools—ChatGPT, Grammarly, Microsoft Copilot, Google Gemini—that their institution has actually licensed, vetted, or blessed for use, as distinct from tools that are merely tolerated or outright banned. CASRAI’s companion guide on how university AI academic-integrity policy has shifted toward tiered permission frameworks covers the policy structure in depth. This page covers the more concrete question underneath it: which named tools and categories of tools actually show up on institutional “approved” lists, why the same tool can be approved for one purpose and prohibited for another, and how to find your own institution’s actual list rather than relying on a generic one.
There is no single, universal “approved AI tools” list
No accreditor, government body, or national higher-education association maintains a master list of AI writing tools approved for academic use. Approval is set institution by institution—usually by some combination of the campus IT security/data-governance office (which vets a tool’s data-handling terms) and the provost’s or teaching-center office (which sets academic-integrity boundaries on how a vetted tool may be used). In practice, the place a student or researcher actually finds their institution’s list is a library research guide or IT-services page, not a university-wide policy document. University library guides at institutions including Cal State LA, the University of Oxford (Bodleian Libraries), Purdue, Florida International University, and National University all publish and maintain exactly this kind of list, updated as tools and licensing arrangements change—which is also why any third-party list, including this one, should be treated as a map to the pattern rather than a substitute for checking your own institution’s current page.
The real distinction: institutionally licensed tools vs. consumer tools
The single factor that most consistently separates “approved” from “not approved” across institutions isn’t the tool’s capability—it’s whether the institution has a contract covering how the vendor handles data.
- Institutionally licensed (enterprise) tools come with a data-processing or enterprise agreement the university’s IT and legal offices have actually reviewed. That agreement typically specifies that prompts and outputs aren’t used to train the vendor’s underlying models and that data handling meets the institution’s compliance requirements (FERPA, and often research-data or HIPAA-adjacent terms for research use). Because of that paper trail, these tools are the ones most often explicitly named as approved, and sometimes as the preferred or required option for any work involving institutional or research data.
- Consumer-facing (free/public) versions of the same underlying AI models usually carry no such institutional agreement. Many institutions still permit their use for low-stakes tasks like brainstorming or grammar checking, but explicitly restrict entering unpublished research data, human-subjects data, or other sensitive or export-controlled material into them, precisely because the vendor’s standard consumer terms may permit that input to be retained or used for training.
This licensing distinction is separate from—and layered on top of—the academic-integrity question of whether AI assistance is permitted on a given assignment at all. A tool can be fully approved from a data-governance standpoint and still be prohibited by a specific course’s syllabus, and vice versa. See the policy-trends guide above for how institutions structure that second, task-level layer, often using a shared reference framework for that task-level permission, as covered in the policy-trends guide.
Tools that recur on institutional approved/licensed lists
The specific named tools vary by institution and change as licensing arrangements are renewed, but the same handful recur across the library guides and IT pages that publish this information:
- ChatGPT Edu — OpenAI’s offering built specifically for universities, deployed institution-wide under a student data privacy agreement rather than OpenAI’s standard consumer terms. Where a university has purchased ChatGPT Edu, it’s typically the version explicitly named as approved, distinct from the free, public ChatGPT that the same institution may restrict for sensitive-data use.
- Microsoft 365 Copilot Chat with enterprise data protection — covered by the same contractual terms a university already has with Microsoft for email and file storage, at no additional licensing cost in many education agreements. Microsoft’s own documentation states that prompts and responses under enterprise data protection aren’t used to train the underlying foundation models, which is the specific property IT offices point to when approving it over the consumer Copilot experience.
- Google Gemini within Google Workspace for Education — the equivalent arrangement for institutions on Google’s education suite, with data handling governed by the Workspace for Education agreement rather than Gemini’s consumer terms.
- Grammarly (Education tier) — widely licensed at the department or institution level for grammar, style, and citation-adjacent assistance. It’s worth noting Grammarly is a general-purpose writing-and-communication platform rather than a tool built specifically for scholarly writing conventions, which is a separate consideration from whether it’s institutionally licensed; CASRAI’s guide to AI tools for improving academic writing style covers that distinction in more depth.
Being institutionally licensed doesn’t automatically make a tool’s output exempt from disclosure. Most policies that approve a tool for data-governance reasons still require the same disclosure of AI assistance—what was used, for what task—that CASRAI’s guide to disclosing AI assistance in a thesis or dissertation covers, and that mirrors the disclosure norms publishers apply to submitted manuscripts.
What “approved” does not mean
An institutionally approved or licensed tool is not the same thing as a blanket permission to use AI on any assignment. Three distinctions matter here, and conflating them is the most common source of confusion:
- Approved for data governance means the vendor’s data-handling terms have been reviewed and accepted by the institution—a security and compliance decision, made centrally.
- Permitted for a given task means a specific course, assignment, or assignment stage allows AI assistance—an academic-integrity decision, usually made at the course or instructor level, and the one CASRAI’s policy-trends guide covers in detail.
- Disclosed means the student or author has stated, per whatever format the course or journal requires, that and how AI assistance was used—a documentation requirement that typically applies regardless of whether the tool itself was approved.
A tool being on an institution’s approved-vendor list answers only the first question. Treat the other two as separate checks, not assumptions that follow automatically from the first.
What’s commonly restricted or banned outright, regardless of institutional licensing
Even at institutions with a broad approved-tools list, some uses are consistently prohibited across the publicly documented policies:
- Submitting substantially AI-generated text as one’s own original work where the course or assignment requires original authorship and doesn’t disclose or authorize that use.
- Entering unpublished research data, human-subjects or patient data, personally identifiable student information, or export-controlled material into a consumer-facing AI tool that lacks an institutional data-processing agreement—this restriction applies regardless of whether the tool is otherwise “approved” for general use, because the concern is the specific data entered, not the tool’s academic-integrity status.
- Commercial contract-writing or “AI essay mill” services that produce complete, submission-ready work on request—these are typically treated the same as any other form of contract cheating, independent of whether they use AI.
How to find your own institution’s actual list
Because approval is set locally, the reliable path is always to check your own institution rather than a generic list:
- Library research guides. Most university libraries maintain an actively updated AI-tools or AI-literacy guide listing which tools are licensed and what they’re recommended for—this is usually the most current and most concrete source.
- IT security or data-classification pages. These specify which tools have passed a security review and what data classification level (public, internal, restricted) each is cleared for.
- Your syllabus or course policy. This is where task-level permission actually lives, per the tiered-framework pattern covered in CASRAI’s policy-trends guide.
- Your graduate school or research-integrity office, for thesis, dissertation, and manuscript-specific disclosure requirements, which often differ from coursework rules—see CASRAI’s guide on AI tools for dissertation writing.
Related CASRAI reference terms: generative AI, large language model, and the generative-AI disclosure statement.
Frequently asked questions
Is ChatGPT approved by universities?
It depends on the version and the institution. Many universities have licensed ChatGPT Edu institution-wide, with data governed by a student data privacy agreement, and treat that version as approved for coursework and, subject to course policy, research use. The free, public ChatGPT typically remains permitted for lower-stakes tasks at many institutions but is often restricted for sensitive or research data, since it isn’t covered by an institutional data agreement.
Are AI writing tools banned at some universities?
A small number of course-level and, less commonly, institution-level policies still prohibit AI assistance entirely, particularly for graded written work. But the broader trend documented across publicly available policies is away from blanket bans and toward tiered, task-specific permission—see CASRAI’s guide to that shift for the detail.
Does using an approved AI tool mean I don’t have to disclose it?
No. Institutional approval is a data-governance decision about the vendor; disclosure is a separate academic-integrity and authorship requirement that applies to permitted AI use regardless of which tool was used.
Is Grammarly considered an “AI writing tool” that needs approval?
Most institutions with a tiered policy do treat Grammarly-style grammar and style assistance as generally permitted, often without the same disclosure burden as full AI drafting—but policies vary, and some require disclosure of any AI-assisted tool. Check your specific course or journal policy rather than assuming grammar tools are automatically exempt.
