What's new in Azure AI Search

Learn about the latest updates to Azure AI Search functionality, docs, and samples.

Note

The release date and content listed below only correspond to actual deployment of the Azure Public Cloud in most cases.

This article provide the evolution history of Azure service on Azure Public Cloud for your reference that may not be consistent with the actual deployment on Azure operated by 21Vianet.

Note

Preview features are announced here, but we also maintain a preview features list so you can find them in one place.

December 2025

Item Description
2025-11-01-Portal migration for agentic retrieval Knowledge bases and knowledge sources in the Azure portal have been updated to use the 2025-11-01-preview REST APIs instead of the 2025-08-01-preview. Portal-created knowledge bases now support the retrieval reasoning effort, and query-time properties (maximum runtime and maximum output size) have been removed from the UI. The portal continues to support search index and blob knowledge sources only; other knowledge source types must be created programmatically.

If you previously created knowledge bases or knowledge sources in the portal, those objects still use the 2025-08-01-preview schema. For help with breaking changes, see Migrate your agentic retrieval code.

2025 announcements

Month Type Announcement
December Portal updates Knowledge bases and knowledge sources in the Azure portal have been updated to use the 2025-11-01-preview REST APIs instead of the 2025-08-01-preview. Portal-created knowledge bases now support the retrieval reasoning effort, and query-time properties (maximum runtime and maximum output size) have been removed from the UI. The portal continues to support search index and blob knowledge sources only; other knowledge source types must be created programmatically.

If you previously created knowledge bases or knowledge sources in the portal, those objects still use the 2025-08-01-preview schema. For help with breaking changes, see Migrate your agentic retrieval code.
November REST API Search Service 2025-11-01-preview is a new preview REST API version providing programmatic access to the data plane operations described in this table.
November Agentic retrieval Semantic ranker and agentic retrieval on free tiers. Both semantic ranker and agentic retrieval are now available on free tiers in select regions, subject to limits on query volume.
November Agentic retrieval Knowledge agents renamed to knowledge bases. Knowledge sources remain unchanged. This rename introduces breaking changes to REST API routes and properties. For help with migration, see Migrate your agentic retrieval code.
November Agentic retrieval Knowledge bases (preview) support new retrievalInstructions and outputConfiguration properties for improved query planning and execution control. It also provides a new reasoning effort for control over LLM processing.
November Agentic retrieval Execute retrieval operations with support for reasoning effort, zero-model-call mode for efficiency, and partial responses.
November Portal update Portal support for knowledge sources and knowledge bases (preview). Use the Azure portal to create and manage knowledge sources and knowledge bases, with a new chat playground for sending retrieval requests. These portal-generated objects use the 2025-08-01-preview schema and have breaking changes with the 2025-11-01-preview. For help with migration, see Migrate your agentic retrieval code.
Foundry IQ (preview) New integration that allows agents in Foundry Agent Service to invoke knowledge bases in Azure AI Search. Foundry IQ offloads complex retrieval operations to the knowledge base, enabling the agent to provide accurate, citation-backed responses based on enterprise data and web sources.
November Skills Azure Content Understanding skill (preview) wraps Azure Content Understanding in Foundry Tools to extract structured Markdown from text, images, PDFs, Microsoft PowerPoint, Microsoft Word, and more. This skill provides advanced document parsing with better table extraction (including cross-page tables), image descriptions, and semantic chunking. For indexed knowledge sources, this skill is available through the contentExtractionMode property within ingestionParameters.
November Security Elevated read permissions for ACLs (preview). New capability to assign elevated read permissions to administrators for investigating problems with ACL configurations used in document access control.
November Security Document-level sensitivity label indexing (preview). New integration with Microsoft Purview to sync document sensitivity labels to the index, honoring their labels and protection at query time for data governance.
November Queries Scoring function aggregation (preview). New capability to combine and aggregate multiple scoring functions, enabling more sophisticated relevance customization and weighted signal combination.
November Queries Facet aggregations (preview). New facet aggregation operations, including minimum, maximum, average, and cardinality, provide enhanced analytics in faceted search experiences.
November Endpoints azure-api.net endpoint support (preview). The Azure OpenAI Embedding skill and Azure OpenAI vectorizer now accept azure-api.net endpoints for Azure API Management (not custom endpoints).
November Endpoints services.ai.azure.com endpoint support. The GenAI Prompt skill, Azure OpenAI Embedding skill, Azure OpenAI vectorizer, and AI enrichment now accept services.ai.azure.com endpoints for Microsoft Foundry resources. When you upgrade from Azure OpenAI to Foundry, a new project is automatically created and becomes available for RAG and multimodal RAG in the Import data (new) wizard.
September REST API Search Service 2025-09-01. New stable REST API version supports general availability for Microsoft OneLake indexer, Document Layout skill, and other APIs.
September General availability Document Layout skill.
September General availability Normalizers.
September General availability Index description.
September General availability Rescoring of binary quantized vectors.
September General availability Rescoring options for scalar compressed vectors.
September General availability Scoring profiles for semantically ranked results.
September General availability Truncate dimensions.
September General availability Unpack @search.score to view subscores in hybrid search results.
September Portal update The Azure portal is undergoing a three-phase rollout to unify the two import wizards. For Phase 1, the Import and vectorize data wizard has been renamed to Import data (new) and redeveloped to support keyword search, modernizing the legacy Import data workflow with an improved interface and user experience.

Import data (new) isn't a direct replacement for the old wizard. For example, it supports fewer skills for keyword search.

Both wizards are currently available, but Import data will be deprecated in a future phase.
September Security Support for Azure confidential computing. Configure confidential computing during service creation to process data in use on confidential VMs. This compute type isn't intended for general use, but rather for stringent regulatory, compliance, or security requirements.

Confidential computing adds a 10% surcharge to the base cost of billable tiers. For more information, see the pricing page.

Now generally available through the 2025-05-01 version of Services - Create or Update (REST API) and the Azure portal.
August REST API Search Service 2025-08-01-preview. New preview REST API version providing programmatic access to the data plane operations described in this table.
August Agentic retrieval Knowledge agents (preview). Architectural changes to the knowledge agent definition, which now requires one or more knowledgeSources instead of targetIndexes and deprecates defaultMaxDocsForReranker. New retrievalInstructions and outputConfiguration properties provide greater control over query planning and execution. For help with breaking changes, see Migrate your agentic retrieval code.
August Agentic retrieval Knowledge sources (preview). New REST APIs for creating and managing knowledge sources, which represent content that knowledge agents use to ground and answer queries. In this preview, you can create knowledge sources for search indexes and Azure blobs.
August Agentic retrieval Answer synthesis (preview). New answerSynthesis modality for knowledge agents. When specified, an LLM generates a natural-language answer as an embedded step in the retrieval pipeline. This differs from the default extractiveData modality, which returns raw search results for downstream processing.
August Agentic retrieval "Fast path" for knowledge agents (preview). (Removed in the 2025-11-01-preview. This documentation no longer exists). The attemptFastPath boolean in knowledge agents enabled a shorter processing time if queries are concise and the initial response is sufficiently relevant. Replacement feature is the minimal retrieval reasoning effort.
August Agentic retrieval Retrieval instructions (preview). New retrievalInstructions property for knowledge agents guides query planning in an agentic retrieval workflow. For example, you can specify criteria for including or excluding specific knowledge sources.
August Indexers Improved indexer runtime tracking information (preview). Applies to Standard 3 High Density (S3 HD) services only. Get Service Statistics response now provides cumulative indexer processing information for the entire service. Get Status - Indexers provides the same information, but for a specific indexer.
August Vector search Strict postfiltering for vector queries (preview). New strictPostFilter mode for the vectorFilterMode parameter. When specified, filters are applied after the global top-k vector results are identified, ensuring that returned documents are a subset of the unfiltered results.
August Vector search Increased maximum dimensions for vector fields. The maximum dimensions per vector field are now 4096. This update applies to all stable and preview REST API versions that support vectors and doesn't introduce breaking changes.
May Agentic retrieval Agentic retrieval (preview) creates a conversational search experience powered by large language models (LLMs) and your proprietary data. Agentic retrieval breaks down complex user queries into subqueries, runs the subqueries in parallel, and extracts grounding data from documents indexed in Azure AI Search. The output is intended for agents and custom chat solutions. A new knowledge agent object is introduced in this preview. Its response payload is designed for downstream agent and chat model consumption, with full transparency of the query plan and reference data. To get started in the portal, see Quickstart: Agentic retrieval.
May Vector search Multivector support (preview). Index multiple child vectors within a single document field. You can now use vector types in nested fields of complex collections, effectively allowing multiple vectors to be associated with a single document.
May Queries Scoring profiles with semantic ranking (preview)​. Semantic ranker adds a new field, @search.rerankerBoostedScore, to help you maintain consistent relevance and greater control over final ranking outcomes in your search pipeline.
May Security Document-level access control (preview). Flow document-level permissions from blobs in Azure Data Lake Storage (ADLS) Gen2 to searchable documents in an index. Queries can now filter results based on user identity for selected data sources.
May Multimodal search Multimodal search (preview). Ingest, understand, and retrieve documents that contain text and images, enabling you to perform searches that combine various modalities, such as querying with text to find information embedded in relevant complex images. See Quickstart: Search for multimodal content for portal wizard support and Azure AI Search Multimodal RAG Demo for a code-first approach.
May Skills GenAI prompt skill (preview). A new skill that connects to a large language model (LLM) for information, using a prompt you provide. With this skill, you can populate a searchable field using content from an LLM. A primary use case for this skill is image verbalization, using an LLM to describe images and send the description to a searchable field in your index.
May Skills Document Layout skill (preview). New parameters are available for this skill if you use the 2025-05-01-preview API version or later. New parameters support image offset metadata that improves the image search experience.
May Portal update Import and vectorize data wizard enhancements. This wizard provides two paths for creating and populating vector indexes: Retrieval Augmented Generation (RAG) and Multimodal RAG. Logic apps integration is through the RAG path.
May Index Index "description" support (preview). The latest preview API adds a description to an index. Consider a Model Context Protocol (MCP) server that must pick the correct index at run time. The decision can be based on the description rather than on the index name alone. The description must be human readable and under four thousand characters.
May REST API 2025-05-01-preview. New data plane preview REST API version providing programmatic access to the preview features announced in this release.
July REST API Search Management 2025-05-01. Stable release of the REST APIs for the control plane operations described in this table. For migration guidance, see Upgrade to the latest REST API in Azure AI Search.
July General availability Pricing tier change. Now generally available through the sku property in Update Service (REST API) and the Azure portal.
July General availability User-assigned managed identity assignment. Now generally available through the identity property that associates a user-assigned managed identity with a search service configuration. Only the assignment step, via the Update Service (REST API) or the Azure portal, is generally available. APIs used for data source or model connections that include a user-assigned managed identity are still in preview.
July General availability Network security perimeter. Now generally available through the Azure Virtual Network Manager REST APIs, which are used to join a search service, and the Search Management REST APIs, which are used to view and synchronize the configuration settings. Portal support for both steps is also generally available.
March Pricing Pricing tier change (preview). Change the pricing tier of your search service. This provides flexibility to scale storage, increase request throughput, and decrease latency based on your needs. Initially, this preview only supported upgrades between Basic and Standard (S1, S2, and S3) tiers, but starting in July 2025, it supports upgrades and downgrades between these tiers. Available in Update Service (2025-02-01-preview) and the Azure portal.
March Queries Facet hierarchies, aggregations, and facet filters (preview). New facet query parameters support nested facets. For numeric facetable fields, you can sum the values of each field. You can also specify filters on a facet to add inclusion or exclusion criteria. Available in Search Documents (2025-03-01-preview) and the Azure portal.
March Vector search Rescore vector queries over binary quantization using full precision vectors (preview). For vector indexes that contain binary quantization, you can rescore query results using a full precision vector query. The query engine uses the dot product of the binary embeddings and the vector query for rescoring, which improves the quality of search results. Set enableRescoring and discardOriginals to use this feature, and call the latest preview API version on the request.
March Queries Semantic ranker prerelease models (preview). Opt in to use prerelease semantic ranker models if one happens to be available in your region. Available in Create or Update Index (2025-03-01-preview).
March REST API Search Service REST 2025-03-01-preview. Preview release of REST APIs for data plane operations. Adds support for multi-vector embeddings, hierarchical facets, facet aggregation, and facet filters.
March REST API Search Management 2025-02-01-preview. Preview release of REST APIs for control plane operations. Adds support for in-place upgrade to higher capacity partitions, in-place upgrade to higher tiers, and Azure Confidential Compute.
April Demo RAG Time Journey. Code and video demonstrations of Retrieval Augmented Generation (RAG) workflows that use Azure AI Search. Segments include fundamentals, patterns and use-cases, vector indexing at scale, and agentic search where you use an agent to evaluate a result and generate a better answer.
February Security Customer-managed keys support for Managed HSM. Use either Azure Key Vault or Azure Key Vault Managed HSM (Hardware Security Module) to store customer-managed keys for extra encryption of sensitive content.

Previous year's announcements

Service rebrand

This service has had multiple names over the years. Here they are in reverse chronological order:

  • Azure AI Search (November 2023) Renamed to align with Azure AI services and customer expectations.
  • Azure Cognitive Search (October 2019) Renamed to reflect the expanded (yet optional) use of cognitive skills and AI processing in service operations.
  • Azure Search (March 2015) The original name.

Service updates

You can find service update announcements for Azure AI Search on the Azure website.

Feature rename

Semantic search was renamed to semantic ranker in November 2023 to better describe the feature, which provides L2 ranking of an existing result set.

  • Last updated on