Get AI summaries of any video or article — Sign up free
Best tools for organizing PDFs in Obsidian thumbnail

Best tools for organizing PDFs in Obsidian

Nicole van der Hoeven·
6 min read

Based on Nicole van der Hoeven's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.

TL;DR

Obsidian can deep-link to PDF pages inside a vault, but it’s weak at categorizing non-markdown PDFs and isn’t a complete solution for PDF annotation/processing.

Briefing

A 127GB pile of RPG rulebooks and reference PDFs was slipping through a highlight-and-notes workflow built around ebooks, because PDFs mix “fiction” text with “reference” material that doesn’t fit neatly into typical annotation flows. The solution that finally matched the creator’s needs is Zotero—used as the PDF library and metadata engine—then wired into Obsidian so PDFs become searchable, taggable, and clickable down to specific pages.

The requirements were practical rather than academic. Citations mattered little; the priority was organizing by tags and categories (so RPG elements like character options, adventures, and bestiaries could be queried later in Obsidian via Dataview), plus “processing” that still allows reading and annotating on mobile or other devices. Just as important were two reference features: deep links that open a PDF at the right page from inside Obsidian, and the ability to keep copies locally and in the cloud (notably Dropbox) while syncing across macOS, Linux, iPad, and Android. Sharing with other gamers also mattered.

Trying to do everything inside Obsidian ran into hard limits. Obsidian handles references well when PDFs are stored as vault assets—hash-based links can open a PDF at a specific page, enabling a table-of-contents style experience. But PDFs aren’t easily categorized because they aren’t markdown notes, and PDF reading/annotation support is not robust enough without extra plugins.

Several alternatives failed on key constraints. DEVONthink could tag and deep-link, but annotation couldn’t be carried into Obsidian, and syncing with Dropbox/iCloud triggered repeated indexing and file-state errors. It also pushed the workflow toward “local-only,” and its mobile companion (DEVONthink To Go) is iOS-only, making cross-device use with Android unrealistic. Readwise performed strongly for highlighting and exporting highlights into Obsidian, but it wasn’t designed as a PDF repository: tagging within the PDF library was effectively read-only, deep links weren’t reliable, and sharing PDFs wasn’t a core strength.

A macOS-focused approach—Skim plus file tagging—worked for lightweight deep links and quick annotation, but it tied the system to macOS and didn’t handle cloud-only PDFs well. That meant online-only files wouldn’t be consistently available, and sharing/tag portability across different operating systems remained weak.

Zotero, by contrast, hit the full checklist, with one tradeoff: the integration relies on Zotero plugins rather than Obsidian plugins. The workflow uses Zotero libraries/collections to categorize RPG PDFs, then uses plugins to import folders automatically and to copy Zotero tags into Obsidian. For page-level reference, Zotero context-menu tools generate special “select item” links; with the right identifier syntax and page numbers, Obsidian links can open the exact PDF location. To avoid duplicating 127GB of files, Zotero is configured to use linked attachments pointing to a base directory (such as Dropbox). Finally, Obsidian becomes the database layer: tags and metadata flow in, Dataview queries surface structured lists (e.g., Arcadia magazine sections), and clicking a result jumps straight to the relevant page inside the PDF.

The end result turns “forgotten folders” into an actively usable RPG reference system—one that scales across devices and stays maintainable even outside an academic citation mindset.

Cornell Notes

The core problem was organizing a massive library of RPG PDFs (127GB) that didn’t fit an ebook-style highlight workflow. Obsidian can deep-link to PDF pages inside a vault, but it struggles with tagging/categorizing non-markdown PDFs and with robust PDF annotation. Zotero becomes the organizing layer: it supports libraries and collections for RPG categories, tagging for later querying, and local+cloud handling when attachments are linked rather than copied. With a couple of Zotero plugins, tags and annotations can be exported into Obsidian, and Zotero-generated “select item” links can be formatted so Obsidian opens the PDF at a specific page. The payoff is a searchable, clickable RPG reference database powered by Zotero metadata and Obsidian’s Dataview.

Why didn’t an Obsidian-only approach work for organizing RPG PDFs?

Obsidian handles PDF references well when PDFs are stored as vault assets: adding a hash-based page reference lets a link open the PDF at a specific page, which supports table-of-contents style navigation. But categorization is weak because PDFs aren’t markdown notes, so tagging and later querying (e.g., via Dataview) is awkward. Processing and annotation also aren’t robust enough without additional plugins, which undermines the “one system” goal.

What were the key requirements used to judge tools, and which ones mattered most?

The workflow prioritized non-academic needs: (1) categorize via tags and later querying in Obsidian (Dataview), (2) process by reading/annotating on devices without breaking the files, and (3) reference through deep links to a specific page plus the ability to save/share PDFs across local and cloud storage. Citations were largely irrelevant. Sharing with other gamers and syncing across macOS, Linux, iPad, and Android were also major constraints.

What broke down with DEVONthink for this use case?

DEVONthink could tag and deep-link, and it worked best when everything stayed local. It failed on “processing” because annotations couldn’t be brought into Obsidian. It also produced many errors when PDFs were stored on Dropbox or iCloud, requiring re-indexing and causing file-state problems. Cross-device syncing was further constrained because DEVONthink To Go is iOS-only, while the creator relies heavily on Android.

Why did Readwise fall short even though it’s strong at highlights?

Readwise is excellent for highlighting and exporting highlights into Obsidian using an official plugin, and it stores everything in its cloud. But it wasn’t built to function as a PDF repository for this workflow: tagging/category management for PDFs wasn’t available (read-only in the reader), deep links to exact PDFs/pages weren’t reliable, and sharing PDFs wasn’t a core capability. The result was good highlight processing without the organization and page-level navigation needed for RPG reference material.

How does Zotero integrate with Obsidian to deliver page-level deep links and tagging?

Zotero organizes PDFs into libraries/collections and supports tagging inside Zotero. Plugins enable two key bridges: (1) Folder import to bring in large sets of PDFs automatically, and (2) Zutilo to copy Zotero tags to the clipboard so they can be pasted into Obsidian. For reference, Zotero’s context-menu link tools (configured in Zotero preferences) generate “select item” links; using the correct identifier syntax and adding the target page number allows Obsidian markdown links to open the PDF at the intended page.

Why does Zotero use “linked files” instead of storing copies, and how is that configured?

With 127GB of PDFs, duplicating files would waste storage and complicate syncing. Zotero is configured with a linked attachment base directory (e.g., a Dropbox folder) so imported PDFs are referenced via links rather than copied into Zotero’s own storage. When moving to another computer, updating the base directory path keeps the links working as long as the underlying files are the same set synced through the cloud.

Review Questions

  1. What specific Obsidian limitation makes tagging and Dataview querying of PDFs difficult, and how does Zotero address it?
  2. Which tool(s) failed primarily due to cloud-sync errors, and what symptoms were described?
  3. How do “linked attachments” in Zotero change storage behavior compared with storing copies, and why does that matter for a 127GB library?

Key Points

  1. 1

    Obsidian can deep-link to PDF pages inside a vault, but it’s weak at categorizing non-markdown PDFs and isn’t a complete solution for PDF annotation/processing.

  2. 2

    A successful RPG PDF system needs tags for later querying (e.g., Dataview), page-level deep links, and cross-device syncing across macOS, Linux, iPad, and Android.

  3. 3

    DEVONthink’s local-first design and repeated indexing errors with Dropbox/iCloud make it a poor fit when PDFs must live across cloud storage and multiple devices.

  4. 4

    Readwise excels at highlight workflows and exporting highlights to Obsidian, but it lacks practical PDF tagging and reliable deep links to exact PDF locations.

  5. 5

    Skim plus macOS file tagging can provide fast deep links and lightweight annotation, but it’s too tied to macOS and doesn’t handle cloud-only PDFs well.

  6. 6

    Zotero is the best match when used as the PDF library/metadata layer, with Obsidian acting as the database interface for tags and Dataview queries.

  7. 7

    Using Zotero “linked attachments” (via a linked attachment base directory like Dropbox) prevents duplicating a massive PDF collection and keeps links working across computers.

Highlights

Zotero becomes the organizing layer, while Obsidian turns that metadata into a searchable, clickable RPG reference database.
Page-level deep links work by generating Zotero “select item” links and formatting them with the correct identifier syntax plus a target page number.
The biggest storage win comes from Zotero linked attachments—pointing to PDFs in Dropbox instead of importing copies into Zotero’s own storage.
DEVONthink’s cloud-sync behavior caused repeated errors and re-indexing, and its mobile app support didn’t match an Android-heavy setup.
Obsidian-only organization breaks down because PDFs aren’t markdown notes, making tagging and Dataview-style querying awkward.

Topics

Mentioned