The No results agent in AI Assist analyzes searches that return no results and recommends ways to improve search performance. It identifies potential issues with search settings, searchable attributes, and synonyms, then suggests targeted fixes. Apply supported changes directly in AI Assist or follow guidance to address other issues, helping more users find relevant results.

No more jumping between two rule editors. You can now build every type of rule right in the Visual Editor - context-only rules, remove or replace a word, replace a query, return custom data, and apply search parameters - along with matching facet placeholders for the Filter and Boost matching attributes consequence. Together with conditionless rules, which landed a few months ago, that means the full set. Set up your rules and preview exactly how they'll behave, all in the place you already merchandise. Click here to learn more.

Query Suggestions can now provide more relevant suggestions when shoppers search within a specific category, department, brand, seller, or other scope. Configure a browsing facet in the Dashboard to use your existing catalog structure and show suggestions that match the shopper’s selected scope. This makes experiences such as “Search within this category” easier to build and more relevant. Click here to learn more.

A/B testing now supports Bayesian experimentation as an additional method for analyzing experiments. Bayesian results show the probability that a variant performs better and provide credible intervals on the likely uplift of business metrics. Customers can choose between Bayesian and frequentist methods when creating a test and receive guidance when results indicate a Bayesian test is ready to stop.

When a pinned product becomes unavailable (because it went out of stock for example), Algolia previously filled that position with an organic search result, breaking the merchandiser's intended curation order. Now, we have introduced a new toggle “Shift pins to close gaps” in the Visual Editor when pinning which fixes this: enable the toggle on a rule and continuous pins will form a ‘pin block’ - when one pinned item inside the block becomes unavailable, the others shift up automatically. Off by default on existing rules, on by default for new ones. You can have many pin blocks or one big pin block. Learn more in this Algolia blog post and review the additional documentation here.

The Algolia AI Search & Discovery app for Shopify now supports version 5 of the JavaScript API client. Some stores have been automatically migrated from v4 to v5 with the compatibility layer. You can review or change your current search client version under Search options → Search client version to either v5 (recommended), v5 with compatibility layer (which runs v5 while keeping most v4 calls working), or v4 (deprecated).

If your storefront uses custom JavaScript that calls the Algolia client directly, we recommend switching to v5 with compatibility layer first, then updating deprecated v4 calls using the JavaScript upgrade guide.

The compatibility layer is scheduled to be removed on February 14, 2027. Stores using custom v4 client calls should update their code and switch to v5 before then. JavaScript client v4 reaches the end of Algolia support on August 14, 2026, after which Algolia Support can no longer provide support for v4-specific issues.

Merchandisers running time-sensitive campaigns - flash sales, Black Friday and Cyber Monday promotions, midnight launches can now select a timezone directly in the date range picker when scheduling a rule. The Dashboard defaults to your browser's detected timezone and handles UTC conversion automatically on save, so promotions start and end exactly when intended. Available in the Visual Editor.

Understanding your analytics is key to improving search performance, but finding issues or opportunities can take time.

AI Assist can now answer questions about your search analytics within the Algolia dashboard. Looking for a general overview? Ask Assist to summarize your search performance, highlight recent trends, or identify opportunities for improvement. Or, if you want to drill down into a specific question, ask about individual metrics or compare performance across indices or time periods.

To get started, open Assist and click “Analyze my search performance”, or ask your analytics question in natural language.

If you need to work with your data outside the dashboard, Assist can also generate downloadable CSVs containing the analytics you've requested, making it easy to continue your analysis offline.

DocSearch UI library v5 introduces an improved search experience powered by Agent Studio, with MCP-based tools, prompt suggestions, and conversational memory. Users can also refine results with facet filters, breadcrumbs, and badges. Customers upgrading from v4 should review the migration guide, as some Ask AI parameters have changed. Click here to learn more. 

DocSearch MCP is now available in public beta, enabling MCP-compatible AI coding assistants to search current public developer documentation and answer with cited sources. No signup, API key, or authentication is required to use the hosted endpoint. This helps reduce outdated or unsupported answers by grounding responses in current documentation and making sources easy to verify. Click here to learn more.