AI for research on the sources you already have
Use AI on sources you already hold, never as a source itself: upload the papers, ask for the claim, method and limitations of each, then ask where two sources disagree. Every quotation and figure must be checked back against the document before you rely on it.
The dangerous way to use AI in research is to ask it what the literature says. It will answer fluently from memory, and some of that answer will be invented.
The safe and genuinely fast way is to attach the documents and confine the model to them. ChatOTL is built for that: multi-file uploads, one thread per topic, branching for tangents and a keyword search across your own message history.
A grounded research loop
- Attach the paper and ask for claim, method, sample and stated limitations.
- Ask what the paper does not establish — the gap between its result and the headline.
- Attach a second paper and ask specifically where the two disagree.
- Ask it to argue against the conclusion you are drifting towards.
Working across a set of documents
Up to 5 files per message at 15 MB each: PDFs, images, plain text, Markdown, CSV, JSON and source code. Text and code are extracted inline; PDFs and images are passed to the model through short-lived signed URLs. Keep one thread per research question so the context compounds.
Find what you already asked
Search your own message history by keyword to recover the summary you wrote three weeks ago instead of regenerating it — and getting a slightly different answer.
Verification is not optional
Ask the model to point to the section or page a claim comes from, then open the document and check it. Anything it cannot locate in an attachment should be treated as unsupported.
Where it fits
Triage a reading pile
Summarise each paper's claim and method to decide what deserves a full read.
Compare sources
Attach two papers and ask exactly where their conclusions conflict.
Stress-test a thesis
Ask for the strongest objection to the argument you are building.
Example prompts
- From the attached paper only: claim, method, sample size, and the limitations the authors state.
- These two papers reach different conclusions. Where exactly do they diverge, and on what evidence?
- Argue against my summary. What would a hostile reviewer say?
Limitations
- ChatOTL cannot browse the web, run searches or retrieve papers — you supply every source.
- It can produce plausible citations that do not exist; never accept a reference it was not given.
- Summaries flatten nuance, so anything load-bearing should be read in the original.
Frequently asked questions
Can it find papers for me?
No. There is no web access and no live search. It works only from the documents you attach, which is the reason its answers stay anchored to real text.
Will it invent citations?
Any language model can. The mitigation is to confine it to attached sources and ask it to point at the page or section behind each claim so you can check it.
How many documents can I work with at once?
Five files per message, 15 MB each, and you can keep attaching across messages in the same thread as the conversation develops.
Is unpublished research private?
Row-level security scopes conversations, messages, settings and uploaded files to the account that created them, and you can use your own provider keys or endpoint.
Start with the paper on your desk
Attach it, ask for claim and limitations, then check the page references.
Open ChatOTL