Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryProductsNew to AI

Semantic search

Also written: Semantic Retrieval, Semantic models, Semantic context, Search indexes, Index

Searches based on the meaning and context of queries, not just keywords.

In plain terms

Semantic search understands what you mean when you search, not just the exact words you type. It uses AI to figure out the intent behind your words.

For example, if you search for "best place to get coffee near me," semantic search can understand you're looking for a coffee shop, not just pages that happen to contain the words "best," "place," "get," "coffee," "near," and "me." It can then show you relevant coffee shops, even if their websites don't use those exact phrases.

The expert version

Semantic search employs AI models, often large language models or specialized embedding models, to represent queries and documents as numerical vectors in a high-dimensional space. Similarity between these vectors, typically measured by cosine similarity, indicates semantic relevance. This contrasts with keyword search, which relies on exact word matching or boolean logic.

This approach allows for understanding of synonyms, related concepts, and the overall intent of a query, leading to more accurate and comprehensive search results. Tradeoffs include computational cost for embedding generation and retrieval, and potential biases inherited from training data.

Why it matters now

Recent advancements in embedding models from Google Gemini are making semantic search more powerful by mapping diverse data types like text, images, and audio into a unified space. This allows AI systems to understand and retrieve information based on meaning and context, rather than just keywords. Microsoft is also integrating this capability into Copilot, connecting it to enterprise data for more accurate and grounded business answers.

Where it shows up 25 items, newest first
How each company uses it
Google Gemini
Google Gemini is developing new embedding models that map text, images, and other data types into numerical representations, enabling faster and more scalable semantic search and recommendations.
Microsoft Copilot
Microsoft Copilot is connecting to enterprise data using semantic context, allowing it to provide accurate answers grounded in business information.
xAI Grok
xAI Grok is mentioned in relation to semantic search, indicating its use of this technology.
Related terms

Vector databaseEmbeddingsLarge Language ModelsInformation retrievalNatural Language ProcessingKeyword search

Often confused with
Keyword search
This method matches exact words or phrases in a query to documen
Fuzzy search
This technique finds strings that are approximately equal to a giv
Boolean search
This method uses operators like AND, OR, and NOT to combine keyw
Origin

The concept of semantic search has roots in information retrieval research dating back decades, but its modern implementation is heavily driven by advances in natural language processing and machine learning, particularly with the rise of deep learning models for embeddings.

Try this about 5 minutes

Use a search engine like Google or Bing and try searching for a concept using different phrasing. For instance, search for "how to fix a leaky faucet" and then try "dripping tap repair." Notice how the results are similar because the search engine understands the meaning behind both queries. Many AI chatbots also allow you to ask questions in natural language, demonstrating semantic understanding.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

How is semantic search different from regular search?
Regular search, or keyword search, looks for exact matches of your words. Semantic search understands the meaning and context of your words to find more relevant results.
What makes semantic search work?
It uses AI, like language models, to convert your search query and the content it searches into a numerical format (vectors) that captures meaning. It then finds content with similar meanings.

First seen on Subvolts Mar 25, 2021. Updated Oct 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.