Google spent August testing a version of Search where the AI answer isn't a feature at the top of the page. It's the page. One confirmed desktop experiment puts three buttons under the homepage search box, each dropping the user straight into AI Mode. Another test, reported at the end of the month, takes the expanded AI Overview and shifts users into AI Mode by default, follow-up box already open, conversation already started. The Google AI Mode default test story is still a test story, and that distinction matters. But tests are how Google rehearses defaults, and this rehearsal deserves your attention now, not after the rollout post.
What The Tests Actually Show
Get the facts straight first, because the coverage blurred them. Three separate August moves point one direction. On August 10, Google confirmed the desktop homepage test with AI Mode entry buttons. Through late August, AI Overviews visibly grew to fill more of the results page on some queries. And in the same window, testers documented AI Overviews that hand off into AI Mode automatically rather than waiting for a tap.
None of this is a launch. Google runs thousands of experiments, and this one collides with real counter-forces: the company took its first Digital Markets Act fine this summer, nearly 300 French publishers filed complaints over AI Overviews, and an unambiguous cohort of users actively hates the feature. Spend five minutes in the right subreddits and you'll find guides for stripping AI results out of Google, extensions built for it, and complaint threads running to hundreds of comments. Defaults that annoy users and regulators simultaneously tend to arrive slowly.
But direction and speed are different questions. On direction, Google has been unambiguous all year: homepage entry points, bigger Overviews, AI Mode booking flows for travel, and a June Search Console report that formally splits AI impressions from organic. You don't build reporting infrastructure for an experiment you plan to abandon.
The Math If The Default Flips
Here's the arithmetic that makes this test worth more attention than most. Tracking puts AI Mode sessions at roughly 93% zero-click. Regular Google search already ends without a click 68% of the time, and when an AI Overview is present, click-through to the top organic result drops 58-79% depending on whose study you read. BrightEdge measured Overviews triggering on about 48% of tracked queries back in February, before the August expansion.
Chain those numbers. Today's default search hands organic results a shrinking but real click stream. AI Mode's economics round that stream down to nearly nothing. A default flip wouldn't dent organic traffic; it would swap the distribution model underneath it, from "rank, get clicked" to "get retrieved, get cited, maybe get remembered." The zero-click trajectory has pointed here for two years. The default test is Google checking whether users will accept arriving.
AI search traffic grew 835% this year. Is your content ready?
RankControl generates 26 content formats optimized for ChatGPT, Claude, and Perplexity. Published on your domain, matched to your brand.
Which SaaS Queries Are Exposed First
The exposure isn't uniform, and triaging your own keyword set takes an afternoon.
Informational queries go first. Definitions, how-tos, "what is X" content: a complete synthesized answer removes the click's entire purpose. If your blog's traffic concentrates here, your charts feel the default flip hardest and earliest.
Comparison and alternatives queries follow. AI Mode answers "X vs Y" and "best X for Y" with synthesized tables assembled from whoever it retrieves. The click may die, but the citation becomes the prize, because being the source named in that table is what puts you on the shortlist.
Branded and navigational queries hold. Someone searching your product's name wants your site, and no answer engine replaces the destination. High-intent pages like pricing and login keep their traffic, which quietly makes branded demand, the thing entity building creates, the most defensible traffic you own.
One measurement warning while you triage: Search Console's generative AI report shows impressions only, no clicks and no queries. You can see that AI surfaces displayed you, and nothing else. The visibility that matters inside those answers, who gets cited and named, has to be tracked directly, engine by engine, because Google's own tooling won't tell you.
What An AI Mode Answer Is Built From
One mechanical detail changes how you prepare, and Google has documented it: query fan-out. AI Mode doesn't answer the query the user typed. It generates a set of related queries behind the scenes, retrieves results for each, and synthesizes across all of them. Google's own example turns "how to fix lawns" into parallel retrievals about herbicides, prevention, and chemical-free removal before any answer renders.
The planning consequence is real. Page-per-keyword targeting underperforms in a fan-out world, because the answer assembles from sources across the whole topic neighborhood. A site with genuine topical depth gets multiple chances to be retrieved per answer; a site with one optimized page gets one lottery ticket. When you rebuild an informational page for the AI era, brainstorm the five to ten queries the fan-out will likely spawn and make sure your cluster, and ideally the page itself, covers them.
Fan-out also explains why the exposure math above isn't hopeless. Every synthesized answer needs sources, plural, and it draws them from deeper in the index than position one. Pages that never won the ranking war still get cited when they're the best answer to one of the hidden sub-queries. The answer engine reshuffled the tournament; it didn't close it.
Built by the team that got cited in 48 hours.
Content generation, backlink building, AI visibility tracking, and Google rankings. One platform, zero guesswork.
How SEO Changes If This Ships
Funny enough, the strategy answer isn't "abandon SEO." Google's own guidance says AI Overviews and AI Mode run on the same core ranking systems as Search, and retrieval strength still decides who's in the answer's candidate pool. Rankings stop being the scoreboard and become the qualifying round.
What changes is where the game is won. Extraction-ready structure decides whether the answer engine can lift your claims. Entity strength and independent mentions decide whether it trusts you enough to name you. And measurement moves to citations per engine, because the August reshuffles proved answer sources can reorder overnight while rankings barely move. Teams keeping a weekly citation baseline saw those shifts in days; everyone else diagnosed them from traffic reports a month later, backwards.
The honest preparation list is short: triage your keyword exposure by the three buckets above, rebuild your highest-value informational pages for citation rather than clicks, put real effort into the branded-demand flywheel, and get the per-engine baseline running before any default flips. All of it compounds whether or not this particular test ships, which is what makes it the rare Google-test story with a no-regret response. You can run that loop by hand, or RankControl's agents handle the content and the weekly tracking while you keep the calls.
The Test Is The Message
Watch for three tells on timing: whether the homepage buttons survive into fall, whether the AIO-to-AI-Mode handoff expands beyond the test cohort, and whether Q3 earnings language starts describing AI Mode engagement the way it once described mobile. Even if this exact test dies, Google just told you what it wants Search to become. The teams that treat the test as the roadmap get a head start measured in quarters.
See your first AI citation report in under 5 minutes.
No setup calls. No onboarding meetings. Connect your domain and see where AI mentions your brand right now.
