Smart Lookup: semantic search in Obsidian

Smart Lookup lets you search your vault by meaning (embeddings), not exact words.

Use it when you want "notes that are about this" even if the note does not contain your search terms.

Fast decision: Lookup vs Obsidian search vs Connections

Use this when you just want the right tool immediately:

If you want the "current note anchored" workflow, see
Exploring the Connections view.

The Smart Lookup loop

When Smart Lookup is the right tool

Smart Lookup shines when you are dealing with classic PKM friction:

• Fuzzy recall: you remember the idea, not the phrase you used.
• Vocabulary drift: you wrote it months ago with different words than you would use today.
• Conceptual research: you want examples, counterexamples, or related sub-ideas.
• Avoiding duplicate work: you want to check if you already solved something before re-solving it.
• Starting a draft: you want to quickly gather your prior thinking before you write.
• Decision recall: you want "where did I decide X?" or "what tradeoffs did I list?"

If you need exact matching (quotes, operators, tags, filenames, regex), use Obsidian's built-in search.

Lookup vs Connections: which view should you open?

Both features surface semantically related notes, but the "anchor" is different:

• Connections view: "What is related to the note I am looking at right now?"
• Smart Lookup: "What in my vault is related to this question or idea I typed?"

A simple rule:

• If you are already in a note and want nearby context, use Connections:
  Connections list feature.
• If you are starting from a question (or you are between notes), use Smart Lookup.

Opening the Lookup view

You can open Smart Lookup from:

• The ribbon icon (left sidebar)
• The command palette
• A custom hotkey you assign in Obsidian

In the command palette, search for:

• Smart Connections: Open: Lookup view

Tip: If you use Lookup daily, assign a hotkey in Obsidian:
Settings -> Hotkeys -> search "Smart Connections: Open: Lookup view".

Writing better queries

Smart Lookup is not a keyword syntax language. Treat it like asking your vault a question.

A simple query formula

Most high-quality lookups include:

• Topic
• Context nouns (domain, project, people, tools, constraints)
• Desired output (checklist, examples, counterarguments, steps, decision, summary)

Example:

• "My notes about topic for context nouns - give me a desired output"

Patterns that work well

• Describe what you want to find
  • "Notes about preventing information overload while researching"
• Add context nouns
  • "For ADHD, distraction, notification fatigue, digital minimalism"
• Ask for a specific kind of help
  • "What are my best checklists for X?"
  • "What are the tradeoffs or counterarguments to Y?"
• Use your current problem
  • "Why do I keep over-capturing info and how did I handle it before?"
• Refine by iteration
  • Start broad, then add 1-2 constraints based on the first results.

"Bad query -> better query" rewrites

• Too vague:

  • "productivity"

• Better:

  • "Notes about reducing context switching while writing"
  • "My approaches to batching reading and processing highlights"

• Too general:

  • "meetings"

• Better:

  • "My recurring meeting prep checklists and follow-up templates"
  • "Notes about meeting decisions and how I record action items"

• Too many jobs at once:

  • "Summarize my notes about sleep, exercise, diet, and focus and give me a plan"

• Better:

  • "What did I write about sleep and focus tradeoffs?"
  • Then: "What habits did I recommend for better sleep consistency?"

Example queries you can copy

• "Ways to reduce information overload without losing important inputs"
• "My notes about boundaries for news/social media consumption"
• "Strategies for batching reading and processing notes"
• "How to keep a PKM system simple when it starts getting complex"
• "Where did I write about focusing on outcomes instead of collecting sources?"
• "What did I decide about X and what reasons did I give?"

Understanding the results

Results are sorted by similarity to your query.

What you will typically see:

• The query (highlighted in blue) is used as the reference point for the results
• A score (underlined)
  • higher means "more related" to the query
• The note name follows the connection score
  • may include a note heading or line range to reference a specific section of the note
• Expandable items (expand or collapse) that let you preview relevant content before opening

How to use the score:

1. Treat it as a ranking signal, not a grade.
2. Compare scores within the same lookup, not across different queries.
3. If the top score is low, your query is probably too vague (add context nouns).

Sources vs Blocks (granularity)

Lookup can return either:

• Sources (whole notes)
• Blocks (more granular chunks)

If results feel too coarse (you keep opening notes and hunting inside them), blocks can help.
If results feel too noisy or heavy, sources are often faster and clearer.

See:

• Smart Connections settings
• Lookup lists settings

Semantic vs lexical search (why results can feel surprising)

That is expected because semantic search is about meaning, not matching text.

Common reasons an exact-match note does not show up:

• The note mentions your words in a different sense than your intended meaning.
• The note is long or mixed-topic, so the embedding represents broader content.
• Another note explains the same idea more directly (even with different wording).
• Your vault is still indexing/embedding new or changed content.

When you truly need "find this exact phrase", use Obsidian search.
When you want "find notes like this", use Smart Lookup.

Troubleshooting

"My exact note did not appear"

Try:

• Add 2-3 distinctive context words (domain, project name, key terms you used at the time).
• Query the concept, not the title.
• Switch from broad ("productivity") to specific ("notification fatigue", "batch processing").

If you still suspect something is wrong, check:

• Whether the note is excluded from Smart Connections indexing
• Whether your embeddings/index are up to date
• Your display settings (Sources vs Blocks, limits, filters)

See the settings guide for deeper controls:

• Lookup lists settings

"Results feel noisy or too broad"

Try:

• Add constraints: "for writing", "for research", "for meetings", "for ADHD", etc.
• Ask for a narrower output: "checklist", "steps", "framework", "examples"
• Reduce the scope: one question at a time

If noise persists, tune the granularity and limits in:
Smart Connections settings.

Related pages

• Getting Started with Smart Connections
• Exploring the Connections view
• Smart Connections settings
• Smart Connections overview
• Smart Context Clipboard
• Smart Chat