Extract links from a PDF offline by opening the PDF Link Extractor in your browser, choosing a local PDF of up to 25 MiB and 40 pages, and starting extraction without uploading the file. The tool reads clickable PDF link annotations rather than scanning every string that resembles a URL, so its report reflects destinations PDF.js exposes as supported annotations. Results stay on your device and appear in page order, with external and internal targets shown as inert text alongside page numbers and link types. You can copy the report or download it as UTF-8 TXT or formula-safe CSV. The process handles one PDF at a time; encrypted, damaged, unsupported, or over-limit documents return a visible error. External links are included only when their scheme is HTTP, HTTPS, mailto, or tel. Unsafe or unsupported schemes are skipped rather than made clickable. This workflow suits link inventories, migration checks, document QA, and reviewing where a PDF directs readers, but it does not open destinations, test whether they work, or decide whether they are safe.

Why annotations matter for offline extraction
For this exact task, use the PDF Link Extractor rather than a generic URL text scanner. The PDF and extracted targets remain on your device, and Lizely does not upload the document or send its contents to a link-analysis server. Here, “offline” describes local document processing; it does not mean the extracted addresses are visited or verified.
The extractor uses PDF.js annotation data instead of searching raw PDF bytes for strings that look like URLs. The PDF.js getAnnotations documentation provides the page-level annotation reference for this approach. That distinction reduces false positives from compressed streams, font data, metadata, and ordinary printed text. It also explains why a visible address can be absent from the report: visible text and a clickable annotation are separate PDF structures.
| What appears in the PDF | Report treatment | Important boundary |
|---|---|---|
| External link annotation | Included when the scheme is HTTP, HTTPS, mailto, or tel | Shown as inert text; it is not opened or checked |
| Internal destination exposed by PDF.js | Labeled internal and given a bounded, readable representation | Not guaranteed to resolve to a final page number |
| Visible URL-like text without a link annotation | Not extracted as a link | Text and annotations are different PDF structures |
| JavaScript, data, file, blob, control-character, or other unsupported target | Skipped and counted | The target is never made clickable |
External scheme checks are case-insensitive, and targets containing control characters are rejected. Internal targets require different interpretation. A named destination or an explicit destination array may point to another location in the same document without containing a web address. The report labels it as internal and serializes a bounded, readable representation. It deliberately does not promise a final page number because destination structures, page references, and viewer behavior can vary.
Prepare a PDF within the tool's limits
Choose a local document and confirm that it meets every preparation requirement before starting the extraction job.
| Preparation check | Accepted value |
|---|---|
| Selection | One local PDF |
| File size | Up to 25 MiB |
| Page count | Up to 40 pages |
| Document state | Unencrypted, readable, and supported |
If a condition is not met, including when the file is encrypted, malformed, unsupported, or over a size or page limit, the tool returns a visible error. It also enforces bounded limits on annotation count, target length, and total report characters so that one document cannot create unbounded work or output.
PDF.js and its shared worker load only after you press the extraction button and remain outside the initial page bundle. The document is still processed locally, and its contents are not sent to a link-analysis server.
Extract links and save a local report
- Choose one local PDF. Confirm that it is no larger than 25 MiB, contains no more than 40 pages, and is unencrypted. Keep the original document available as the source record while you review the extracted report.
- Click Extract Links. The tool requests each page's link annotations through PDF.js, selects supported external or internal destination fields, rejects unsafe schemes and control characters, and preserves page and annotation order. Exact duplicate targets on the same page are removed, while the same target on different pages remains visible. Every displayed target stays inert.
- Review and export the report. Read each target with its page number and link type, then check the summary for examined annotation counts, accepted links, duplicates, and skipped targets. Copy the plain-text report or download a formula-safe CSV or UTF-8 TXT file. The export contains the same inert target data shown on screen.
Read the ordered results correctly
The report preserves the document's reading sequence. Page order comes first, and annotations remain in their original order within each page. Each supported target is associated with its page and link type, while CSV output also provides an annotation number.
- Page order: Results move from one page to the next, making it easier to connect a destination with the part of the document that contains it.
- Annotation order: Multiple links on one page remain in the sequence PDF.js exposes rather than being rearranged alphabetically.
- Duplicates: An exact duplicate on the same page is removed, but a repeated destination on another page stays visible because its page occurrence may matter during review.
- Internal targets: These are labeled as internal and may appear as a named destination or a readable destination representation rather than a conventional web address.
A repeated destination across different pages is not merely redundant. Its page occurrence can show where a reference appears and help distinguish separate uses of the same link. Treat internal labels as destinations exposed by PDF.js, not as verified page jumps.
Export TXT or CSV for later review
Choose the export format according to how you will use the results. CSV is suited to structured review in spreadsheet software, while TXT provides portable plain text.
- CSV: The report contains page, annotation number, type, and target columns. Cells beginning with spreadsheet formula prefixes are neutralized before serialization, while quotes and line breaks are escaped using CSV rules.
- TXT: The download is UTF-8 plain text and uses stable page labels, making it suitable for archiving, searching, or pasting into another local workflow.
- Copy report: Copying uses the same safe, inert report provided by the extraction result rather than converting targets into active links.
Formula neutralization reduces one CSV-specific risk, but it does not make document-derived content trustworthy. Treat every extracted target as untrusted data, especially before forwarding the report, sharing it, or pasting it into another system.
Reasons an expected link may be missing
If an address you expect is absent, consider what kind of record the PDF actually contains before treating it as an extraction failure.
- The URL is printed as visible text but was not made into a clickable link annotation.
- The external target uses a scheme outside HTTP, HTTPS, mailto, and tel, or it contains rejected control characters.
- The destination is internal and represented by a named or explicit destination structure rather than a web URL.
- The URL-like string occurs in metadata, font data, a compressed stream, or ordinary text without a supported annotation attached to it.
- The whole document was rejected because it was encrypted, malformed, unsupported, or outside the file-size or page limit.
For an external item, a missing row can mean PDF.js did not expose a supported link annotation for it, or that an exact duplicate target on the same page was removed. That absence does not prove the PDF contains no textual mention of the address.
Use the inventory without treating it as a safety check
Use the extracted inventory for link inventories, migration checks, document QA, or reviewing where a PDF sends readers. Page numbers remain useful during these tasks because the same destination may appear in several places. Keep the original PDF unchanged as the source record, and use the TXT or CSV export as a separate working document.
The report is not a crawler, broken-link checker, phishing detector, accessibility audit, or content scanner. The tool does not navigate to, fetch, validate, reputation-check, or guarantee the safety of an extracted destination. Its presence in the report means only that PDF.js exposed a supported link annotation. Verify important destinations independently before visiting or publishing them.
Related reading: PNG to PDF on Chromebook: A Local File Workflow.
Related reading: Extract Images From PDF Without Losing Quality: JPG or PNG.