Apple is working on a new generative search engine integrated into Siri that it internally calls World Knowledge Answers, “response engine” capable of offering condensed and contextual results. The initiative, reported by Bloomberg sources, points to a launch window that's closer than one might expect for a recently created project.
In spite of Initial stumbles with Apple Intelligence and the classic assistant architecture, the company has rebuilt the project on large language models. The current plan places the arrival of this experience in March 2026, with a debut within Siri and subsequent expansion to other areas of the system.
What is Apple's World Knowledge Answers (WKA)
World Knowledge Answers is the internal name of the system with which Apple seeks to compete with proposals such as Perplexity or the Google AI summariesThe idea is that the user formulates a query and receives a brief and clear response, prepared by Siri from information from the web and the device, with support for text, photos, videos and local points of interest.
The more solid forecasts They point out that the WKA will be part of a major Siri update in iOS 26.4 (March 2026)At launch, the feature will be integrated into Siri and won't yet appear in Spotlight or Safari, two destinations Apple plans to bring it to later.
This schedule comes after a deep reconstruction of the assistant: the first generation version did not offer the expected reliability and Apple decided to completely remake it with a second-generation architecture based on LLMThe company maintains an internal model comparison (a "bake-off") to decide which specific technology will power each part of the system, without affecting the target date.
How the new system works
The technical basis of the response engine is based on three blocks that work together to interpret the request, search for information, and generate a clear result. This organization allows Siri to offer useful summaries and citable without forcing the user to jump between multiple links.
- Planner: interprets the query (voice or text) and decides the steps.
- Search module: combines device content and web results.
- Summarizer: prepares a short, structured response for the user.
In addition to handling multimedia content, Siri will be able to extract data from the context on the screen and, when the user allows it, rely on personal information to answer questions with greater relevance.
AI models and potential partners
Apple evaluates own and third-party models for various parts of the system. According to recent reports, the company has signed a formal agreement to try Gemini, Google's model, with a view to supporting summary functions in Siri. In parallel, it continues testing solutions from Anthropic (Claude) and internal developments.
The company has made it clear that, when it comes to processing personal data (emails, messages or other sensitive content), it will exclusively use its Foundation models already running on infrastructures under its control, reinforcing its focus on privacy.
Who is behind the project
The effort involves several teams: the Siri group that reports to Craig Federighi (software), the AI division led by John Giannandrea and the services area led by Eddy Cue, with the participation of profiles such as Mike Rockwell (Vision Pro) into key pieces. Under the umbrella of the Siri Renewal —including internal projects referred to as “Linwood” and the LLM evolution— lays the foundation for the knowledge search engine.
In model training, Apple had been using mostly synthetic data; now it compares that material with real user data that choose voluntarily, with controls designed to minimize risks and biases, which should improve the understanding and practical usefulness of the responses.
What you can do with the response engine
With the new Siri, you can ask general questions and get a concise summary using web sources, ask for nearby references (e.g., restaurants or museums), or combine queries that include on-screen content and authorized personal data to find information faster.
- Open questions with condensed results and reference links.
- Queries with images, videos or nearby places of interest.
- Actions chained through apps, guided by voice.
- Cross-web and cross-device searches to provide more context.
