Keep a term of AI coursework, and be able to show it
Study tools each keep their own history and none of them keep yours. One archive across all nine, organised by module, with the record your department asks for.
For students running a term of work through several AI tools who have to be able to account for it.
The situation
A term of coursework now runs through three or four different tools. NotebookLM for the reading, ChatGPT or Claude for working through a problem, Perplexity or Gemini for finding something to cite. Each keeps its own history, and none of them keep yours.
Come assignment week you need two things at the same time. The material you worked out five weeks ago, and an honest account of how you used AI to get there. Neither one survives in a chat sidebar full of threads named after whatever the model guessed the topic was.
Departmental policy has moved faster than the products have. Most now ask you to declare how you used AI, and some ask for the prompts. Reconstructing that in April from work you did in February is close to impossible, so people either guess or leave it vague.
What Savelore does about it
- The study tools other extensions skip
- NotebookLM, Google AI Studio, Perplexity and DeepSeek are supported alongside ChatGPT, Claude, Gemini, Grok and Copilot. Most extensions in this space cover the first four and stop, which leaves out the ones a lot of reading and source-finding actually happens in.
- Organised by module, not by platform
- Folders and tags by module, assignment or reading week. What platform something came from is already recorded, so your own labels can carry the thing the software cannot know.
- Collection mode for a reading week
- Pull the passage that mattered out of four separate sessions, across different tools, into one entry with numbered citations back to each source.
- The model, the date and your prompt
- Every entry records the platform, the model and when you saved it, and collection mode lets you keep your own prompt beside the answer. Those are the pieces a disclosure statement is made of.
- A link a supervisor can open
- Publish an entry as a public page and a tutor, a supervisor or a study group can read it without an account or an install. Unpublish it after.
- Usable without paying
- Saving, folders, tags, filters, source links and publishing are all on the free plan, which has caps on how much you can save rather than on what you can do.
How to keep a term's work in order
Set this up in the first week of term and it costs you a few seconds per save after that.
- 1
Make a folder per module
Not per platform. You will be looking for the essay, not for the tool you happened to use that day.
- 2
Save the thing that changed your understanding
Not every response. The one where the explanation finally landed, or where you got a source worth chasing.
- 3
Keep your prompt with it
Use collection mode to hold your question next to the answer. A response without the question it answered is much weaker evidence of anything.
- 4
Write one line about what you took from it
In your own words. This is what you will search on, and it is also the raw material of the disclosure you write later.
- 5
Build the reading week into one entry
When a topic spans several sessions across several tools, collect the passages into a single entry with citations rather than four separate saves.
- 6
Write the disclosure from the archive
Filter by module and read off what you actually did, with dates and models attached, instead of trying to remember it.
Reconstructing your AI use in April does not work
The honest version of a disclosure statement needs specifics: which tool, roughly when, and what you asked it for. Nobody remembers that two months later, so what gets written is a vague sentence that satisfies nobody and would not survive a question about it.
An archive with dates and models on it turns that into reading rather than remembering. Filter to the module, look at what is there, write what happened.
The tools most extensions leave out
A lot of study work happens in NotebookLM and Perplexity, and a lot of the free-tier experimenting happens in Google AI Studio and DeepSeek. Extensions in this space overwhelmingly cover ChatGPT, Claude, Gemini and Grok, which misses those entirely.
Savelore covers all nine, and an entry from NotebookLM sits in the same archive with the same tags and filters as one from Claude.
Citing and disclosing are different jobs
Citing is what you do when you quote or paraphrase something a model produced. Disclosing is the account of how you used AI across the whole piece, which most departments now ask for separately.
The archive helps with both, and the details of citation formats are covered on the page about keeping a citable record.
The honest limits
Savelore does not tell you what your institution permits, and policies vary a lot between departments, let alone universities. Check yours.
It does not format citations for you either, and it will not make an AI-generated claim reliable. It keeps a record you can stand behind, which is a smaller and more useful job than it sounds.
Frequently asked questions
Does it work with NotebookLM?
Yes, along with Google AI Studio, Perplexity, DeepSeek, ChatGPT, Claude, Gemini, Grok and Copilot.
Is it free?
The free plan covers saving, folders, tags, filters, source links and publishing an entry as a public page, with caps on how much you can save. Pro removes the caps and adds export and Notion sync.
Will it write my citations?
No. It keeps the model, the date, the source link and the text together so you have what a citation needs. Formatting it is still your job.
Can my tutor see my archive?
Only what you choose to publish. Publishing sends one entry, and nothing else in your archive goes with it. You can take the page down afterwards.
Do I need an account?
Not to save. An account is needed to publish an entry, to turn on cloud sync, and for Pro. Each of those is separate from the others.
Is my work stored on someone's server?
Not by default. Entries live on your own machine unless you turn cloud sync on.
The work you did in week three is worth having in week eleven.
Most of a term's AI use evaporates because it lives in four separate histories nobody goes back to. Savelore keeps what mattered, organised by module, with the dates and models still attached.