For libraries, graduate schools and IT units

Turn your students' AI use into a disclosure

As your students write, AI use is recorded paragraph by paragraph, and the disclosure draft is built from that record. The student sees the record; the institution sees only aggregate counts.

Grongy's institution access report

A detection score does not show how a thesis was made

Writing is spread across tools

A student searches for sources in one place, keeps notes in another and writes in a third. Where each sentence came from is recorded nowhere.

A convincing reference can be invented

Chat models can produce a reference from memory. Because it looks right on the page, neither the student nor the advisor can tell at a glance.

A policy is hard to turn into practice

The guidelines ask for an AI disclosure, but the student can't remember what to disclose. The statement is written from memory just before submission.

What the institution sees, and what it does not

There is no institutional login. It isn't a setting that can be switched off: there is no institution account to open, and so no student document to open. The institution cannot see through the record; a student can still give their advisor a link (below).

The institution sees

  • How many accounts joined, and the seat cap
  • How many accounts were active in the last 7 and 30 days
  • How many AI requests were made since the start
  • The share of use by area: chat, writing, review, sources, analysis, other
  • Joins by week

Below five members, the report shows only that count; the breakdowns stay hidden. The report never holds a name, an email address or any text. It opens from a link we give you and is closed to search engines.

The institution cannot see

  • The student's text, drafts and notes
  • The AI-use record
  • The student's library and uploaded PDFs
  • The student's version history and chats

No student text ever accumulates on the institution's side; you judge the pilot on numbers alone.

The advisor sees the draft, not the AI-use record

A student shares their own document through a link and chooses the access themselves.

ViewNo account needed

Whoever opens the link reads the text.

CommentA Grongy account needed

The advisor leaves comments and does not change the text.

EditA Grongy account needed

The advisor's corrections arrive as tracked changes.

A view or comment link does not show the live draft: it shows the version the student last saved with “Update snapshot”. Only an edit link updates on every save. Tracked changes cover typed edits only; pasting and structural edits are not tracked.

The advisor cannot see

  • The AI-use record and the disclosure draft
  • The library and uploaded PDFs
  • The version history

Your students work from source to submission in one place

A library from the scholarly indexes

Papers from OpenAlex, Crossref and DOAJ, plus PubMed and arXiv when the topic fits them, and the student's own PDFs; highlights and notes in one place.

Your department's template

Importing the department's Word template into an empty document takes its page setup. There are dozens of thesis and paper structures, national thesis formats among them; for the reference list, students choose from more than 10,000 journal and university styles.

An assistant bound to sources

Drafting and rewriting work from the project library; a search result that gets cited is added to the library. A citation with no match in the library does not enter the generated text. Chat can also search the web.

Statistics without code

Tests reported in APA format, and a qualitative coding workspace. The numbers come from computation; the model writes only the interpretation.

Checks before submission

Citations are compared against the sources' own text. Retracted papers, unregistered DOIs and mismatched records are flagged without using the allowance. There is also a fixed submission checklist.

No tool for getting past detection

There is no feature for making text slip past plagiarism software or AI detectors, and our terms of use prohibit using it that way. Grongy does not measure similarity either; we can't say in advance what a tool like Turnitin will show.

You set the scope, we open the record

We open the institution record; there is no self-serve purchase. A pilot is an institution record with a set duration and a seat cap.

  1. Set the scope

    An institution record holds:

    • A name
    • A join code
    • Optional email domains
    • One plan: Plus or Pro
    • A start and an end date
    • A seat cap (can be left open)
  2. Students join through the link

    A student signs in with their own Grongy account and opens the join link. If you give a list of domains, joining is limited to email addresses on them; since email addresses are not verified at sign-up, the code in the link is what really controls who joins. When the seat cap is reached, joining closes.

  3. Judge the result from the report

    The librarian report shows joins and the spread of use in numbers.

On the end date, the accounts return to the free plan. Documents are not deleted and libraries stay in place. A student's own paid subscription, if they have one, continues.

After the end date, the links a student gave their advisor stop working: sharing needs the student to be on a paid plan.

Single sign-on (SSO), SAML and LTI are not supported.

Questions institutions ask

Can we see what our students write?

No. There is no institutional login, so there is no student document to open. The librarian report shows numbers only, and below five members it hides the breakdowns too.

Can we enforce our own integrity policy?

There is no institution-level setting: you can't set a threshold or switch features off. Where your policy applies is the disclosure: the disclosure draft is built from the use record, and the student can edit it to your guidelines. Your policy is what binds; Grongy produces the evidence for the disclosure.

What goes to the AI from uploaded tables and transcripts?

The first 40 rows of a table that is being interpreted, 60 rows when a document summary is asked for, and up to 8,000 characters of a transcript when qualitative codes are suggested.

What happens when the pilot ends?

The accounts return to the free plan; documents and libraries are not deleted. The links a student gave their advisor stop working.

Can we set up single sign-on?

Not at the moment. SSO, SAML and LTI are not supported; students join with their own Grongy accounts, through the join link.

Where is our data kept?

Synced documents are kept in Sweden and the application server runs in Germany; no region is promised for uploaded files. AI requests are processed by our providers, mostly in the United States. Which services are involved and where requests are processed is in our privacy policy.

Do you have a security certification, and is data end-to-end encrypted?

Neither. We hold no ISO 27001, SOC 2 or other security certification. Synced documents and uploaded files are encrypted in transit and at rest, but this is not end-to-end encryption: our service can technically read what is stored.

Let's set up the pilot together

Tell us which program, how many people and how long you'd like to try it, and we'll open the record to match.