A surprising amount of modern research lives at URLs that look public but are still controlled by one person’s account.
A Google Doc linked from a paper, a spreadsheet used to support a news story, a shared folder containing field notes, or a presentation circulated through a conference page may be readable by anyone with the link. From the reader’s point of view, that makes it part of the public web. Underneath, however, it may still depend entirely on the original owner’s cloud account.
That is a fragile arrangement.
Google’s own documentation makes the lifecycle more precise. When an owner moves a file to Trash, other people can continue accessing it until it is permanently deleted; files normally remain in Trash for 30 days unless the owner empties it sooner. Once permanent deletion happens, the public link no longer represents an independently preserved copy.
A public link is not a public copy
This distinction is easy to miss because cloud services deliberately make sharing feel like publishing. You paste a URL into a report, people can open it, and the job appears finished.
But the URL is often only permission to view the owner’s live copy. If that copy is permanently deleted, the reference disappears with it. A small but telling 2025 Google Drive support case involved a user who accidentally deleted a folder named “Public Research Articles” and discovered that recreating a folder with the same name did not restore the old public links. The anecdote is not a survey of scholarly practice; it shows the identity problem plainly. A replacement file is not the file the old URL identified.
The result is a peculiar form of link rot: the address may have been widely cited, but nobody except the account owner necessarily possessed the underlying file.
Exporting changes the preservation model
For research that matters beyond the immediate project, an independent copy is safer than a shared link alone. Google supports downloading Docs and Sheets in ordinary file formats, but a research repository provides a stronger publication model. Zenodo lets researchers publish files with descriptive metadata and assigns a DOI to the resulting record, separating the citable research object from somebody’s working Drive folder.
Exports have limits. A PDF will not preserve a document’s complete edit history. A CSV may lose formulas, formatting, comments, and multiple sheets. Comments and suggestion history can disappear. Some cloud-native behavior has no perfect offline equivalent.
That is why preservation should begin by asking what actually matters: the final text, the data, the revision history, comments, authorship, or the interactive behavior of the cloud document.
Cloud documents are enormously convenient for collaboration. The mistake is assuming convenience creates permanence. A publicly shared research document can be one account deletion away from becoming a citation to nowhere.