Gemini Deep Research comes to Google Finance, backed by prediction market data (arstechnica.com)

Google is adding its Gemini Deep Research capability to Google Finance, rolling out over the next several weeks. The Finance chatbot will gain a “Deep Research” option that runs asynchronous, Gemini-based jobs to produce longer, fully‑cited research reports on complex financial questions — including queries about future outcomes backed by new prediction/betting-market data sources. The experience mirrors the Gemini app’s Deep Research: you submit a prompt, the model compiles evidence and citations, and you return later to view the completed report. Google says everyone can run a limited number of reports; users with AI Pro and AI Ultra subscriptions get higher caps (the Gemini app limits hint at free users getting very few jobs, AI Pro ~20/day and AI Ultra ~200/day). For the AI/ML community this is a notable step in operationalizing retrieval‑augmented generation and model-based research in a regulated domain: it combines large‑model synthesis, citation tracking, and external signal integration (prediction markets) into a product workflow. Technically, expect asynchronous job orchestration, heavier retrieval and provenance pipelines, and higher compute/latency costs compared with simple chat queries. The rollout highlights both opportunities (faster evidence‑backed analysis for traders and researchers) and risks (citation accuracy, overreliance on model outputs, and opaque aggregation of market signals) that will be important to evaluate.