Gemini adds Grounding with Exa for websites, docs, people, and company search
Gemini models can now use Grounding with Exa to search websites, technical docs, papers, people, and companies through Exa's index. That gives Gemini a new agent-style grounding path alongside Google's first-party search tooling.
TL;DR
- ExaAILabs' launch post says Gemini models can now use Grounding with Exa, giving them access to Exa's search index across websites, technical docs, papers, people, companies, and more.
- The official rollout points to both Exa's announcement and Google Cloud's Grounding with Exa docs, which makes this look like a productized Vertex AI integration rather than a one-off demo.
- ExaAILabs' thread opener frames Exa as an "agent-first search" layer for Gemini, which is a different pitch from plain web retrieval.
- Confusion showed up immediately in posts like iScienceLuvr's reaction and ai_for_success's question, because the partner here is Google, whose own search stack is the obvious comparison point.
You can jump straight to Exa's blog post, inspect the Google Cloud documentation, and see the launch framed inside a Google Cloud and Vertex AI context in ExaAILabs' post. The weird part surfaced fast: people immediately asked why Google would plug a startup search product into Gemini at all, as ai_for_success and iScienceLuvr put it.
Grounding with Exa
The core ship is simple. Exa says Gemini models can now call Grounding with Exa, and the source list in that same post names websites, technical docs, papers, people, and companies as searchable entities.
That matters because it is broader than the usual "search the web" framing. Exa is positioning its index as a retrieval layer tuned for agent workflows, not just a generic browser result page, in ExaAILabs' wording and in Exa's announcement.
Vertex AI docs
The launch was not just a social post. Exa linked directly to a Google Cloud documentation page, which suggests the integration is meant to be consumed through Google's existing Gemini and Vertex AI surfaces.
Even the launch image attached to ExaAILabs' post pairs Exa with Google Cloud, Vertex AI, and Gemini Enterprise branding. For engineers, that is the useful detail: this is landing as a supported grounding option inside Google's stack, not as a separate wrapper product.
Google Search still sits next to it
The cleanest answer to the immediate "why would Google need Exa?" reaction came from OfficialLoganK's reply, which says Google Search is still available as a first-party tool in the Gemini API, while Exa joins as an ecosystem partner.
That turns the story from a replacement narrative into a tool menu. The interesting part is not Google outsourcing search wholesale. It is Google exposing another grounding path, one that Exa is explicitly marketing around agent-style retrieval and nontrivial entity search.