AI for research papers — still a reading problem
An AI research assistant that never shows the PDF will happily invent a method. PDF Lens is for the opposite habit: keep the paper on screen, select the dense parts, and ask questions you can cite.
What researchers actually get stuck on
Not “what is this paper about?” — the abstract already tried. The stuck moments are local: a methods clause, a surprising exclusion, a figure caption that assumes a definition from page 4, a related-work paragraph written for insiders.
Those are selection problems. Highlight the clause. Ask PDF Lens to explain it, teach it, or walk through it. The panel docks beside the page so you do not lose the surrounding paragraph.
A working pattern for a paper
- Write the paper’s question in your own words before any AI (how to read a research paper).
- Search the PDF for the outcome, the sample, and the word “limit.”
- Use Paper Chat only for “where do they define X?” and “what do they report for Y?”
- Use Lens on sentences you cannot yet paraphrase.
- Extract evidence as claim + support + page (guide).
Citations you can click
Paper Chat may mark answers with [p. N]. Treat those as invitations to open the page, not as proof. If the sentence is not there, discard the answer. Models miss, and hosted PDF Lens uses an economical model on purpose.
What PDF Lens will not do for you
- It will not OCR a scanned journal scan.
- It will not compare two PDFs in one query.
- It will not replace reading the figures. Retrieval is text.
- It will not give you unlimited tokens. Free and Pro are capped.
If you wanted a notebook over many sources (slides, audio overviews, mixed links), look at how that differs from an AI PDF reader in the NotebookLM note. If you wanted an editor, see Acrobat.
Research reading guides · Paper Chat · Summarizing papers · Pricing