- PDFs are where lookup habits break: the dictionary popup gives senses without context, and copy-paste to a chatbot costs 30–60 seconds a term
- The built-in three-finger tap is fine for plain words; it fails on technical senses ("significant", "degenerate", "attention")
- Selection-based tools work in Preview, Skim, Zotero's reader and browser viewers — anywhere text is selectable
- Scanned PDFs have no selectable text: OCR first, then everything works
- Keep lookups in place: 2–3 seconds next to the sentence beats 30–60 seconds in another window
The short answer
On a Mac, the reliable way to look up a term inside a PDF without losing your place is a selection-based tool that answers in place: select the term, get a short explanation that accounts for the sentence around it, and keep scrolling. The built-in dictionary (three-finger tap) handles plain vocabulary; technical terms need context awareness, which dictionaries don't have.
This post covers the PDF-specific details. For the overall paper-reading strategy, see the complete workflow for reading papers with unfamiliar terms.
Why PDFs are where lookups go wrong
Three frictions are specific to PDF reading:
- The page is fixed. Unlike a web page, a PDF doesn't adapt — no reflow, no reader mode. Whatever you do about a term happens on top of a static page, so every window switch is total: you're out of the document.
- The terms are technical. "Significant" in a results section, "degenerate" in a physics paper, "consideration" in a legal document — the dictionary lists everyday senses first, and the sense you need is third, if it appears at all. We covered this in why the dictionary definition doesn't help.
- The fallback is expensive. The copy → switch to chatbot → paste → read → switch back loop takes 30–60 seconds, and you re-enter the page having lost the sentence's thread. Do that 10 times in one paper and the overhead exceeds the glossing itself.
Your options, honestly ranked
| Method | In place? | Understands context? | Cost per lookup |
|---|---|---|---|
| Three-finger tap (built-in dictionary) | Yes | No — plain senses only | ~2s |
| Select → AI explanation in a bubble (PopGloss) | Yes | Yes — reads surrounding sentence | ~2–3s |
| Copy → paste into a chatbot | No | Yes, if you also paste context | 30–60s |
| Search the term in a browser | No | Eventually, after filtering | 30–90s |
The built-in option is underrated for everyday words and overrated for technical ones. The chatbot option understands context but costs the most attention. The selection-based middle ground is what we compared across Mac lookup tools more broadly.
A workflow that holds up across PDF apps
- Pick a reader with good text selection. Preview and Skim are solid; Zotero's built-in reader is excellent because annotations live with your library. All of them expose selectable text to system-wide tools.
- Select just the term — a word or short phrase. You don't need the whole line; a context-aware tool grabs the surrounding 100–500 characters by itself.
- Get the explanation in place. With PopGloss, a bubble appears next to the selection with a 1–3 sentence explanation plus what the term means in this sentence. Short / medium / detailed lengths are switchable in the menu bar; the academic style is tuned for papers.
- Keep reading. The whole interaction is 2–3 seconds. If the term turns out to be load-bearing — it recurs, or the paper's claim depends on it — park it for a proper deep-dive after the section instead of drilling down mid-sentence.
Scanned PDFs are the one hard limit: image-only pages expose no text to select, so no lookup tool can work there. Run OCR first — most reference managers can — after which selection works normally.
Small habits that compound
- Gloss twice, then decide. A term you've looked up in two different papers is a term worth one real deep-dive. Your lookup history (PopGloss keeps it in a local SQLite database on your Mac, never uploaded) is a ready-made list of those candidates.
- Two-column papers: select the term itself, not the full line. Column order confuses long selections but rarely short ones.
- Split screen is still a switch. A chatbot tiled next to your PDF is better than a hidden window, but your eyes still leave the sentence and the prompt still needs context pasted in. In-place beats side-by-side.
PopGloss is $10 for 1,000 credits that never expire (cost per explanation = model tier × answer length; cached repeats and failed requests cost 0), with 10 free credits to try — no account, no API key. Pricing details if you want the exact math.
Try it yourself: Download PopGloss — 10 free credits included, never expire.