Google puts Gemini 3.5 Flash-Lite behind AI Search

Google is putting a smaller, faster AI model behind its Search answers. On 21 July 2026 it launched Gemini 3.5 Flash-Lite, which it calls its fastest and most cost-effective 3.5-class model, and said the model is coming to Search. The model that writes AI answers is changing under the hood, and that is what decides which brands get summarised and cited.

What happened

Google announced Gemini 3.5 Flash-Lite alongside 3.6 Flash and 3.5 Flash Cyber. It positions Flash-Lite for high-throughput, low-latency work such as agentic search, priced at $0.30 per million input tokens and $2.50 per million output tokens, and says it significantly outperforms the previous Flash-Lite generation on coding, long-context and task-execution benchmarks. Google said the model is coming to Search, where lighter models already help power AI Overviews and AI Mode.

Why it matters

AI Overviews and AI Mode already sit above the classic results for a large share of queries. The model doing the summarising shapes what gets pulled into an answer and which sources it cites. A faster, cheaper model lets Google run AI features across more queries at lower cost, so a growing share of a brand’s visibility is mediated by a machine summary rather than a ranked link. When the model changes, the way your content gets read and quoted can shift with it.

What this means for multi-location brands

For a central team, the surface that decides whether your locations show up in an AI answer is now a moving target, and you do not control which model is behind it. The response is to compete for the answer, not just the ranking: keep location data structured and accurate so an AI can resolve and cite it, and write content that answers the question cleanly in the first lines. Track how the brand appears in AI search results across markets, invest in generative engine optimization, and keep presence consistent at scale through Places AI and coordinated local search marketing. A model swap should not change whether your stores are findable, if the underlying data is solid.

Gemini 3.5 Flash-Lite is priced at $0.30 per million input tokens and $2.50 per million output tokens, positioned for high-throughput, low-latency tasks such as agentic search.

Google

The bottom line

The headline is a model release, but the story for brands is quieter: Google can now afford to put AI answers in front of more searches. Visibility increasingly depends on being the source a model chooses, so the brands that keep their data clean and their answers direct will keep showing up, whichever model is doing the reading.