Baidu overhauls its search engine with AI features and a voice function in its most sweeping revamp in years, and rolls out its first image-to-video model
Context & Ripple Effects
Baidu’s search strategy has been moving toward generative AI since its planned 2023 search integration of Ernie Bot, and its latest redesign makes that shift more central to the product rather than a standalone feature. The company had also recently updated its foundation and reasoning models to compete with rival Chinese AI offerings.
The revamp arrives while Baidu is defending its search business amid intensifying AI competition: its latest quarterly results showed modest revenue growth even as it continued investing in AI. Voice also extends a line of product work that began with its AI assistant for mobile search.
First-order effects
- Baidu users gain AI-led search features, voice interaction and access to the company’s first image-to-video model, expanding the types of queries and media tasks handled within Baidu’s products.
- Baidu turns its search surface into a more prominent distribution channel for its AI capabilities, alongside its recent model upgrades.
Second-order effects
- Chinese AI rivals will face greater pressure to pair model releases with consumer-facing search, voice and multimodal product experiences, not just standalone model performance.
- A more AI-mediated search experience can change how users reach information and services, raising the importance of product integration and default distribution for Baidu.
Third-order effects
- If these redesigns become standard, search competition will increasingly center on multimodal assistants that accept text, speech and visual inputs rather than conventional result-page retrieval alone.
- The shift may deepen the advantage of platforms that already control high-frequency consumer entry points, while making AI model capability and interface design inseparable competitive assets.
The trend: Consumer search platforms are evolving into multimodal AI interfaces, using established distribution to turn foundation models into everyday services.