What is semantic search in Marqly?

Semantic search means Marqly finds your saves by meaning, not by exact keywords. Type “that video about sourdough starters” and Marqly surfaces the right save even if the word “sourdough” never appears in its title. It works because the search looks across everything Marqly knows about a save — the title, the content, your highlights, and video transcripts — and matches what you mean, not just what you typed.

Keyword search vs. searching by meaning

Keyword search fails in a predictable way: it needs you to remember the exact words. If the article was titled “Levain 101: a beginner’s fermentation guide” and you search “sourdough starter,” a keyword engine shrugs. Semantic search understands that levain, fermentation, and sourdough starters live in the same neighborhood of meaning, so your half-remembered description is enough.

That matters because half-remembered descriptions are all anyone actually has six months after saving something. Nobody remembers titles; everyone remembers roughly what a thing was about.

What Marqly searches across

Semantic search doesn’t stop at titles. It works across:

• Titles — the obvious layer
• Content — what the saved page actually says
• Highlights — the passages you marked, often the most memorable part
• Transcripts — for YouTube videos saved with their transcript attached, the spoken words are searchable too

That last one is quietly powerful: a point someone made 40 minutes into a talk becomes findable by describing the point, not by scrubbing through the video.

How it fits with auto-tagging

Tags and semantic search are complementary layers. AI auto-tagging gives you browsable entry points into your library; semantic search handles the “I know I saved this but I can’t remember what it was called” moments — the case folders and keywords were never good at. For practical query patterns and examples, see how to search your bookmarks with AI.

Try it on a real memory

The convincing test is personal: think of something you saved months ago, describe it from memory, and see if it comes back. Try Marqly free — import your existing bookmarks and search them by meaning.