An AI workspace for researchers working across many sources
Research work is long-running and source-heavy, which is what most chat tools handle badly. In ChatOTL a project's sources live in a knowledge base you keep coming back to, threads persist per topic, branching lets you follow a competing interpretation without derailing the main line, and everything is scoped to your account — so confidential or unpublished material stays yours.
Researchers rarely have a single-question problem. They have a corpus, a project that runs for months, and material they cannot paste into a public chat box.
This page is about how those constraints map onto ChatOTL by role. If you want the mechanics of interrogating a document, the research workflow page covers that step by step.
Academics and PhD students
- One knowledge base per chapter or review so a corpus outlives any single conversation.
- A persistent thread per topic keeps your terminology and framing consistent across months.
- Branch to develop a competing interpretation while the main argument stays intact.
Analysts
- Keep recurring reporting packs as a knowledge base and ask what changed against previous framing.
- Attach CSV or JSON for a first structural read before you commit to a model.
- Save a preset holding the questions your team always asks of a new report.
Journalists
- Transcripts, filings and documents in one place, searchable by keyword across your own history.
- Sensitive material stays scoped to your account; you can also bring your own DeepSeek key.
- Ask which documents contradict each other before you build a story on one of them.
Confidentiality, concretely
Row-level security scopes every conversation, message, upload and knowledge base to the account that created it. Nothing is shared between users, and no data is used for other users' answers.
Where it fits
Thesis chapter
A knowledge base per chapter; the corpus survives the semester.
Reading pile triage
Decide which three of ten papers deserve a full read.
Interview corpus
Transcripts in one base, questioned as a set rather than one by one.
Example prompts
- Across my chapter-two knowledge base, which sources support the weaker version of this claim?
- Compare the framing in this quarter's report against the one in my knowledge base.
- Which of these transcripts contradict each other on the timeline?
Limitations
- No literature search: ChatOTL does not browse the web and cannot retrieve papers or verify citations.
- Extracted quotes and figures can be wrong; check each against the source text.
- It has no judgement about quality of evidence — that remains yours.
Frequently asked questions
How do researchers keep a long project organised in an AI tool?
Give each project or chapter its own knowledge base and its own thread. The knowledge base holds the corpus so it is reusable months later; the thread holds the framing so terminology stays consistent.
Is ChatOTL suitable for confidential or unpublished material?
Conversations, uploads and knowledge bases are scoped to your account by row-level security, and are never shared with other users. For stricter control, bring your own DeepSeek key or register your own compatible endpoint.
Can it find papers or sources for me?
No. There is no web access and no live search, so discovery stays with your literature database. ChatOTL works on the sources you bring.
Is it useful for analysts and journalists, not just academics?
Yes — the pattern is the same wherever documents pile up: reports, filings and transcripts in a knowledge base, questioned as a set. The confidentiality guarantees are identical.
Set up your first research project
One knowledge base, one thread, the sources you already have.
Create a free account