"Search a PDF instantly" means two genuinely different things depending on the tool, and knowing which one you're using changes how you should phrase your search.
Keyword search: exact text matching
This is the classic ⌘F search everyone already knows. Type a word or phrase, and the tool finds every place that exact text (or a close variant, depending on case-sensitivity settings) appears in the document.
Strengths:
- Instant, no processing delay — it's just scanning the existing text.
- Precise — if you search "Section 4.2," you get exactly that, no false positives from vaguely related content.
- Works for anything with a fixed, known term: a defined legal term, a specific product name, an exact figure.
Weaknesses:
- Misses anything phrased differently. Searching "termination" won't find a clause that says "either party may end this agreement" — same meaning, no shared keyword.
- Requires you to already know roughly what word was used.
Semantic search: meaning-based matching
This works using the embedding approach described in How AI Understands PDFs — your search query gets converted into a vector representation of its meaning, and the tool finds passages whose meaning is closest, regardless of exact wording.
Strengths:
- Finds relevant passages even when you don't know the exact phrasing used in the document.
- Good for conceptual searches — "what does this say about liability" when you don't know if the document uses "liability," "responsibility," or "obligations."
Weaknesses:
- Slightly slower, since it requires the document to already be processed into embeddings (usually done once, upfront, when you first open it).
- Less precise for exact-term lookups — if you need every instance of a specific defined term, keyword search is more reliable, since semantic search is finding "similar meaning," not "exact string."
- Can occasionally surface a passage that's topically related but not actually what you meant — worth checking the retrieved passage, not just trusting the match.
Comparison table
| Keyword search | Semantic search | |
|---|---|---|
| Finds exact phrase matches | Yes | Not guaranteed |
| Finds differently-worded but related content | No | Yes |
| Speed | Instant | Near-instant, after one-time document processing |
| Best for | Known terms, defined phrases, specific figures | Concepts, questions, unfamiliar terminology |
| Needs the document pre-processed | No | Yes (embeddings generated once) |
Which to use, in practice
- You know the exact word or phrase: keyword search — faster and more precise, no reason to reach for anything more complex.
- You know the concept but not the exact wording: semantic search — this is exactly the gap it's built to close.
- You're not sure: try keyword search first (it's instant); if it comes up empty, that's the moment to try a semantic/AI search instead of assuming the content isn't there.
Searching across multiple PDFs
Both approaches usually start out scoped to one open document. Searching across a whole folder of PDFs — all your course readings, every contract from a client, a full literature review's worth of papers — needs a tool that specifically indexes multiple documents together, not one that just opens each file individually. If this is a regular need, check for it explicitly before assuming a PDF tool supports it; plenty of otherwise-good AI-PDF tools are built around a single open document at a time.
A quick example
Searching a 50-page contract for a specific clause:
- Keyword search "termination": instantly finds every literal use of that word.
- Semantic search "how can either party end this agreement": finds the termination clause even if it's worded as "either party may conclude this arrangement," which a keyword search for "termination" would completely miss.
Using both — keyword first for anything you know the exact term for, semantic for anything you're searching by concept — covers more ground than relying on either alone.