Google’s AI Overviews — the generated answer that now sits above the classic blue links — have gone from an experiment to the default for a large share of searches. To put a number on it, I run a fixed set of everyday queries through Google every week and record how many return an AI Overview, and which sites those overviews cite. This is a companion piece to the AI Overviews Tracker, where the live figures update on their own.
Nearly nine in ten tracked searches now show an AI Overview
At the most recent scan, 88% of the tracked searches returned an AI Overview. That set spans ten industries — health, finance, marketing, legal, home services, software, ecommerce, real estate, education, and general business — so it is meant to reflect a broad mix of everyday intent rather than a cherry-picked slice.
The share is not evenly spread. Commercial and how-to research — the exact queries that used to send steady traffic to businesses — is where AI Overviews are most saturated. Marketing, software, ecommerce, and B2B queries returned an overview essentially every time, while more sensitive categories like legal and personal finance showed them less often, likely reflecting Google’s caution on YMYL (“your money or your life”) topics.
YouTube and Reddit are the most-cited sources by a wide margin
Every time an overview appeared, I recorded the domains it linked as sources. The pattern is striking: YouTube and Reddit dominate, ahead of established reference sites like Investopedia, Mayo Clinic, and Wikipedia. In other words, Google’s AI leans heavily on video and on genuine peer discussion, not only on polished publisher content.
That has a practical implication. If you are trying to be surfaced inside AI answers — the discipline people are calling generative engine optimization, or GEO — a presence on YouTube and in the real communities where your topic is discussed may matter as much as your own site’s pages.
What this means if you rely on search traffic
An AI Overview answers the question before the user reaches your link, which compresses clicks on informational queries. The response is not to panic but to adjust:
- Write to be quoted, not just ranked. Clear, answer-first sections with real structure are easier for a model to lift than a long preamble.
- Earn presence beyond your domain. Given how often YouTube and Reddit are cited, being genuinely useful in those spaces feeds the same engine.
- Shift some measurement from clicks to citations. Being named as a source builds authority even when it does not send a click today.
You can check how ready a specific page is to be cited with the free AI Search Readiness Grader, and watch the overall trend on the AI Overviews Tracker.
How this is measured
Each week, every query in the fixed basket is run against Google (United States, English) through a SERP data provider, and I record whether the page contains an AI Overview and which domains it cites. It is a fixed, curated sample rather than the whole of Google, so read it as direction and contrast between categories, not a population-wide figure. AI Overviews are also personalized and volatile — they vary by location, account, and moment — so the weekly snapshot captures the trend over time, not the guaranteed state of any single search.