100 Questions About Obsidian

Preface

This is a "100 questions" guide: a way to break a large subject into questions, then read it one question at a time.

Obsidian is not just a Markdown editor. It is a local-first knowledge asset. Once AI is connected, the structure, naming, links and acceptance rules of your vault decide whether AI amplifies your thinking or amplifies the mess. The guide follows seven stages: a minimum viable start, daily note habits, AI integration, knowledge-base structure, real workflows, the plugin ecosystem, and the pitfalls.

Quick answers

• Start with the minimum setup and real notes. Do not chase a perfect vault on day one.
• A vault grows: a small stable folder structure, clear naming, and links or MOCs added as needed.
• Make AI answer only from your notes and cite the source. If there is no source, mark it "to confirm".
• Desensitize sensitive material or use a local model, and confirm before anything leaves your machine.
• Let AI organize, retrieve and draft. Keep judgement, opinion and final acceptance for yourself.

Compiled from Obsidian documentation and public material, structured and lightly completed for this site.

This is an edited edition: some source passages were truncated by OCR, so the body focuses on an actionable question list with concise answers.

1. Get started without overwhelm: a minimum viable Obsidian + AI

1. I chat with AI all day and forget it all. Was it wasted?

Only if the conversation leaves no trace. Write the conclusions, decisions and next actions into notes so AI can continue from them next time.

2. I use WeChat favorites as a second brain. Why do I never go back to them?

Favorites are collection, not processing. Turn saved items into notes with a title, tags, a summary and a conclusion so they can be found and reused.

3. People say stronger AI means I don't need notes. Where is the trap?

AI does not have your context or long-term memory. You record your judgement, experience and material; AI is an amplifier, not a replacement for memory.

4. Before asking AI to remember for me, should I decide "for future me" or "for AI"?

Decide the purpose first. For yourself, keep context and judgement. For AI, make it structured, searchable and sourced.

5. I have very few notes. Is it too early to start Obsidian?

No. Start with a minimum setup and real notes. A vault grows from use; you do not need a perfect empty structure first.

6. First time opening Obsidian, what three things should I do?

Create a vault, write one real note, and turn on search and linking. Do not install plugins or build a perfect folder tree yet.

7. Perfect vault screenshots make me anxious. How do I get over it?

Those vaults are outcomes, not starting points. Solve one real problem of your own and let the structure grow around it.

8. What is good about local Markdown, and why does it survive software shutdowns?

It is plain text, portable, readable by AI and not locked to one app. That is the foundation of a long-term knowledge asset.

9. What is the easiest way to sync notes between phone and computer?

Use Obsidian Sync, iCloud, OneDrive or Git. Pick one stable method and avoid editing the same file on multiple devices at once.

10. Folders or backlinks: which should I choose?

Folders answer "where does it live"; backlinks answer "what is it related to". Start with a few folders, then use links for relationships.

11. Is it useful to run plain Obsidian for a week without AI plugins?

Yes. Build the capture habit and the vault structure first. Connecting AI to a messy vault only amplifies the mess.

12. Is the free version enough, or should I buy a membership?

The free version is enough for personal use. Buy Sync, Publish or other add-ons only when you actually need them.

13. Can I give sensitive content (diary, clients) to a cloud AI from the start?

Desensitize it or use a local model first. Give the cloud only the necessary scope and confirm before anything is sent.

14. What is the minimum viable path to experience "AI reads my notes" once?

Create a vault, write ten real notes, install one AI plugin, and ask it a question that can only be answered from your notes.

2. Practical daily-note habits: learn to capture before adding AI

15. I have eight hundred notes and never use them. How do I make notes useful?

A note must be findable, reusable and output-oriented. Give every note a title, a conclusion and a use case.

16. Others link everything and it works; my links make things messier. Why?

More links are not better. Link only notes that are genuinely related, and write the relationship next to the link.

17. What is a MOC, and why is it better than folders in the AI era?

A Map of Content is an index note that gathers notes on one topic. It gives both you and AI a clear entry point, which folders alone cannot.

18. How do I use tags without overusing them?

Use tags for cross-folder topics and status, keep them few and frequent, and do not use tags as a substitute for links.

19. Should I learn Templater?

Learn it if you write the same kind of note often (daily notes, meetings, reading notes). Skip it for occasional notes.

20. How long should one note be? Too long and AI cannot read it.

One note, one topic. If it explains one thing clearly, it is the right length. Split long notes into atomic notes that AI can retrieve.

21. How do I turn backlinks into a knowledge graph AI can use, not noise?

Links need semantics. Write the relationship beside the link, link only relevant notes, and clean up orphan links regularly.

22. How do I keep a daily log from becoming a garbage pile?

Daily logging is fine, but refine it into topic notes regularly. The daily log is raw material, not the finished asset.

23. How do I take reading notes without becoming copy-paste that never gets used?

Do not only copy. Write a summary in your own words, your opinion, quotable lines and action items.

24. What is frontmatter for, and why does AI like it?

Frontmatter gives notes structured properties such as date, tags, source and status. AI and Dataview can both use them.

25. If AI writes my notes, will I stop thinking?

Let AI organize, rewrite and outline; write the judgement and opinion yourself. Otherwise your thinking atrophies.

26. How should I store meeting and interview notes so AI can find them later?

Keep one record and one action list. Include time, participants, conclusions, next actions and sources.

27. How do images, screenshots and PDFs fit into a vault without becoming disconnected?

Put them in a single attachments folder, name them clearly, embed them in notes and add a short description.

28. What is a dumb but reliable naming and folder convention?

Use date plus topic in names, and keep folders few and stable. Do not chase perfect classification; rely on search and links.

29. What should a daily note contain so it becomes future context for AI?

Facts, emotions, decisions, next actions and what you plan to do tomorrow. These are the context AI will use to understand you.

30. My notes are scattered and AI stitches answers together badly. What do I do?

Build MOCs and topic notes, and give AI an explicit retrieval scope and a source requirement.

3. Connect AI to Obsidian: the shortest path

31. First time connecting AI, can I install just one plugin? Which one?

Yes. Start with either Copilot or Smart Connections, get it working, then expand.

32. Can I play without an API key? How do I start with a local model at zero cost?

Yes. Start with Ollama and a local model. The trade-off is slower speed and weaker capability.

33. Copilot or Smart Connections: how do I choose without overthinking?

Copilot is more about chat and vault Q&A. Smart Connections is more about semantic relatedness and discovery. Choose by need.

34. I am afraid of leaking an API key. How do I set up the account safely?

Use a separate account, minimum permissions, and never put the key in notes or screenshots. Use environment variables or the plugin's key store.

35. How do I install a local model (Ollama) in half an hour?

Install Ollama, pull a 7B model, point the Obsidian plugin at the local address, and test a question.

36. The default model fails on a Chinese vault. How do I change embeddings and chat models?

Choose Chinese-friendly embedding and chat models. Test on a small vault first, then scale up.

37. My old computer lacks VRAM. Which small models can I run, and what is the cost?

Run 3B or 7B quantized models, or use local AI only for summaries and retrieval. If that is not enough, use a metered cloud model.

38. How do I make AI answer only from my notes and not invent things?

State: answer only from this vault, say when there is no answer, and cite a source for every conclusion.

39. What should I ask the first time I connect AI to feel that it "knows me"?

Ask something only your notes can answer. That tests whether it actually read your vault.

40. What is the minimum setup for a private, offline AI knowledge base?

Local Ollama, local embeddings and an Obsidian plugin, with both model and vault on your machine.

41. Cloud is strong, local is stable but weak. How do I choose by scenario?

Use local for sensitive or offline work, and cloud for complex reasoning. Split by task.

42. Which plugin is easiest for saving AI conversations into the vault?

Use a plugin that exports conversations, or manually write the conclusions into a topic note.

43. Too many plugins make the vault slow. How do I keep it to two?

Keep two or three core plugins, enable them as needed and remove unused ones regularly.

44. What three guardrails should I set before connecting AI because I fear data leaks?

Desensitize, grant minimum access and prefer local processing. Confirm before anything is sent out.

45. Can a phone run a local AI, or is it only a sync client?

Usually it is only a sync or lightweight client. Heavy models run on a computer or in the cloud.

46. Models change fast. Will today's setup be obsolete tomorrow?

Write the configuration and prompts into a document, choose stable models, and test small upgrades before switching.

4. Turn notes into an asset with AI

47. How do I make AI "understand me" without re-introducing myself every time?

Give it a vault guide, a stable naming structure and an explicit retrieval scope.

48. What is Vault QA, and how do I make AI answer from my accumulated notes?

Make AI retrieve relevant notes first, answer from them, and cite the sources.

49. How do I write a "vault manual" for AI so it behaves?

Describe the structure, naming, topics, forbidden areas, answer format and source requirements.

50. Is my second brain just a journal? How do I tell whether it can "think"?

Check whether it can answer questions, surface connections and support decisions, not just store a log.

51. What is the three-layer raw / notes / wiki structure, and why separate them?

Raw holds original material, notes hold processed thinking and wiki holds concept or topic pages. AI retrieves more cleanly.

52. How do I make AI build the knowledge-base skeleton so I do less?

Give it a topic list and structural rules, and let it generate folders, MOCs and templates for you to adjust.

53. When new material arrives, how do I make AI compile it into the knowledge base?

Use templates, AI summaries and automatic links to put new material in an inbox, then compile it into the wiki regularly.

54. How do I make AI answer from the wiki and cite a source for every conclusion?

Require a note link or title for each conclusion. If there is no source, mark it "to confirm".

55. Should I trust AI's answers completely? How do I catch fabrication on the spot?

Do not. Check the source, date, definitions and original text. Verify key conclusions yourself.

56. How do I make AI give the vault a regular health check?

Every week or month, ask it to find orphan notes, duplicate topics, stale content and missing links.

57. How do I prevent cognitive atrophy when using AI as a second brain?

Let AI organize; you judge. Keep the habit of writing your own summaries and opinions.

58. My notes are too shallow and AI cannot save them. How do I write notes AI can use?

Record conclusions, reasons, examples and action items. The more specific the note, the more AI can help.

59. What is it like when AI actually uses backlinks, and why is linking worth it?

AI can follow links to find context and related notes. Careful linking pays off in retrieval quality.

60. How do I use AI for periodic reviews and weekly reports without letting it think for me?

Let AI summarize facts and draft. Write the opinion and trade-offs yourself.

61. Can AI uncover implicit relationships between notes? How?

Use embeddings or similarity search to find related notes, but judge the relationships yourself.

62. My vault spans years. How does AI use Dataview to index "me over time"?

Add date and topic properties, then use Dataview queries to compare notes across years.

63. How do daily notes connect scattered notes into "me over time"?

Write one daily note, then regularly link relevant entries into topic notes or MOCs.

64. How should I design tags and properties so AI retrieval stays on track?

Keep them few and stable with consistent naming. Tags cover topic and status; properties cover date, source and type.

5. Real scenarios: study, writing, research, work

65. Exam prep: how do I make AI generate practice questions and key points from my notes?

Ask AI to generate questions, identify key points and review mistakes based on your notes. Make it cite the note behind each answer.

66. I forget after memorizing. How do I turn notes into Anki-style flashcards with AI?

Convert notes into question-and-answer cards and use spaced repetition. Let AI draft the cards and check them yourself.

67. When writing stalls, how do I make AI mine topics and material from my vault?

Ask AI to find relevant notes, opinions and quotes, then write the outline yourself.

68. If AI continues a half-written article, will it drift? How do I control it?

Give it the outline, tone and boundaries, and let it write a local section. Keep the final judgement for yourself.

69. My newsletter has no ideas. How does AI generate 30 topics from old notes?

Ask AI to cluster old notes into topic candidates, then choose the ones that fit your audience.

70. Reading papers and reports: how do I make AI summarize and link them to existing notes?

Ask AI to summarize, extract methods and conclusions, and link the new note to related notes you already have.

71. Researchers: how do I read literature closely with local AI without leaking data?

Use a local model and local vault so sensitive material never leaves your machine.

72. Workplace: how do I turn meeting notes into searchable action items with AI?

Extract tasks, owners, due dates and sources, and output a searchable action list.

73. I have too much client and project material. How does AI surface the right note when I need it?

Use consistent naming, tags and MOCs so AI can retrieve by project and context.

74. Writing proposals and reports: how does AI draft from my history without inventing?

Give it the historical material and an outline. Let it draft; verify data and conclusions yourself.

75. Legal, medical and other sensitive industries: how do I use AI without crossing compliance lines?

Desensitize, use local or compliant cloud services, and require human review. Do not fully automate.

76. Annual review: how does AI turn a year of notes into "what I grew into"?

Ask AI to summarize the year, extract changes, results and lessons, and write the judgement yourself.

77. Book bloggers: how do I turn highlights into citable knowledge cards with AI?

Turn highlights into knowledge cards, then extract quotable opinions and cite the source.

78. Learning a language: how does AI create contextual practice from my notes?

Use your notes for scenario practice, mistake cards and retelling exercises.

79. Startups and side projects: how do I manage ideas with vault + AI without losing inspiration?

Use the vault for ideas, validation, customers and reviews; let AI surface related ideas and gaps.

80. I repeat myself. How does AI warn me that I wrote something similar three years ago?

Ask AI to find similar old notes and flag potential repetition before you publish.

81. Cross-team material is scattered. Can a personal vault + AI fill the gap, and where is the boundary?

It can help for personal synthesis, but sensitive and official material needs clear boundaries and permissions.

82. Show me a real workflow: from saved item to output, where does AI do the work?

Saved item → inbox → summary and tags → topic note → output. AI helps with organizing, retrieval and drafts.

6. Plugin ecosystem and advanced play

83. Does Obsidian have an official AI? If not, how do I choose community plugins?

There is no official AI. Community plugins fill the gap. Choose by need, maintenance and permissions.

84. Which two or three plugins are most worth installing in 2026, and which should I avoid?

Copilot, Smart Connections and Ollama-related plugins are common choices. Avoid bloated, over-permissioned or unmaintained plugins.

85. Where do I find templates and starter vaults? Can I copy one instead of building from scratch?

Look at official templates, community starter vaults and GitHub. Try them on a small scale before migrating.

86. Is "backlinks = GraphRAG" hype or genuinely useful? How do ordinary users benefit?

Backlinks provide structure; GraphRAG is retrieval augmentation. It helps, but it is not magic.

87. What are Agentic and MCP, and why are they "for people who want to go further"?

They let AI call tools and read external data. They suit advanced users; start by protecting permission boundaries.

88. What is the minimum viable way to let Claude or ChatGPT read my vault?

Export Markdown, use MCP or a local bridge. Pay attention to privacy and scope.

89. Where do I follow news, tutorials and communities without drowning in them?

Use the official forum, Reddit, GitHub and Chinese communities. Search by problem, not by tutorial feed.

90. Besides the official marketplace, where can I find plugins and themes? How do I install them safely in China?

Official marketplace plus GitHub. Mind network access and source trust; do not install unknown builds.

91. My vault gets slower over time. How do I diagnose and slim it down?

Check plugins, attachments, indexes and models. Use a binary-search approach to isolate the cause.

92. If I want to leave Obsidian, can local Markdown really move with me?

The Markdown moves, but links, plugins and templates need manual adaptation.

7. Pitfalls and guardrails

93. What are the five most common beginner mistakes?

Installing too many plugins, chasing a perfect folder tree, collecting without processing, letting AI think for you, and putting sensitive data into the cloud carelessly.

94. What happens to people who let AI "think for them"?

Their judgement atrophies. Let AI organize and draft; keep the thinking and decisions.

95. After connecting AI, my notes got messier. How do I avoid an "AI garbage dump"?

Build structure and acceptance rules first, then let AI organize. Without them, AI multiplies the mess.

96. What are the three privacy pieces: private data, untrusted content and outbound channels?

Desensitize private data, isolate untrusted content, and confirm every outbound channel.

97. Local models are not zero-cost. How do I diagnose performance problems and set a stop-loss?

Watch VRAM, speed, power and maintenance. Set a stop-loss line; if it fails, downgrade or move to the cloud.

98. Don't let conversations go to waste — but why is "full automation" a bigger trap?

Without human confirmation, automation amplifies mistakes. Semi-automation with checkpoints is safer.

99. The second brain is overhyped. Which promises should I ignore?

It is not memory outsourcing; it is thinking support. Do not believe "one click makes you smarter".

100. What three things do you only understand after a year?

The vault has to grow, AI must answer from your notes, and judgement cannot be outsourced.