Google's AI Overviews now cover 43% of searches — and Search is getting its own agents
Coverage has nearly tripled in a year, AI Mode has passed a billion monthly users, and Google I/O 2026 introduced agents that work inside Search on your behalf.
Alex Rivera
Editor
Google's AI Overviews now appear in roughly 43% of searches, up from about 15% a year earlier, according to data reported in late July 2026. Some measurements from earlier in the year put coverage even higher, in the 48–60% range — the variance reflects methodology, but the direction is unambiguous: AI-generated summaries are now a default part of the search results page for a large share of queries.
AI Mode, Google's more conversational search surface, has grown alongside it — visits rose from 126 million in June 2025 to 279 million by May 2026, and the feature passed one billion monthly users roughly a year after launch, with query volume more than doubling every quarter since.
What it means for publishers
The tradeoff is a well-documented hit to referral traffic: organic click-through rate drops by more than 60% when an AI Overview appears above the results, and around 58% of Google searches now end without any click at all. Early 2026 data suggests a slow partial rebound, but the baseline has shifted — content strategy built solely around ranking position is running into a search page that increasingly answers the query itself.
Search agents
At I/O 2026, Google previewed the next step: Search agents that users can create, customize, and manage directly inside Search, running in the background to reason across information continuously rather than only responding to a single query. Combined with AI Mode's growth, it points toward Search increasingly acting on a user's behalf rather than just indexing pages for them to click through.
More from AI News
OpenAI previews Astra by having it solve ten decades-old math problems for $2,000
An internal version of OpenAI's next major model resolved open questions in group theory, complexity, and combinatorics — with machine-checked proofs published alongside the results.
Alibaba ships Qwen3.8-Max, a 2.4-trillion-parameter model it's open-sourcing next week
The new flagship posts benchmark scores that challenge GPT-5.6 Sol and Claude Fable 5 on agentic and multimodal tasks — and Alibaba says open weights are coming within days.
OpenAI restructures around Sol/Terra/Luna tiers, ships ChatGPT Work
The new GPT-5.6 lineup comes with steep price cuts and an agentic system that runs multi-hour projects across a team's files and apps.