The scale is no longer a test
Google launched AI Mode in the US in May 2025. Twelve months later, it crossed 1 billion monthly active users globally — a pace that took Google Search decades and YouTube around ten years to match.
That number matters to advertisers for one specific reason: AI Mode is no longer a beta surface or an experiment to monitor. It's the primary search experience for a user base larger than Facebook's ad-supported audience. When Google confirmed the milestone, it also released the first detailed behavioral data since launch. Those patterns have direct, practical implications for how your ads perform on a surface that now reaches more people than most social platforms.
How AI Mode users actually search
The gap between standard search and AI Mode isn't a design difference. It's a syntax difference.
In standard Google Search, users type: "running shoes flat feet." In AI Mode, they type: "I'm looking for a running shoe that's comfortable for flat feet and works for half marathons." Google's first-year data confirms this pattern at scale — queries in AI Mode run three times longer than traditional search queries. Users treat the interface like a conversation, not a search box.
The most common first words in AI Mode queries: "what," "how," "I," "is," "can." The presence of "I" in that top five is significant. It means users are expressing personal context alongside the category need — not just stating a query, but framing it around their specific situation. That's a completely different input than the keyword queries your current ad headlines were built to match.
Your seven-word RSA headline was written to fit a 3-word query. AI Mode is a 30-word query environment. The mismatch isn't catastrophic — ads still serve — but the asset surface area that can connect with those longer, conversational queries is much smaller when your copy is compressed to keyword format.
The "which" problem for purchase decisions
The fastest-growing query category in AI Mode is searches beginning with "which" — growing 40% faster than overall AI Mode usage. Users aren't searching for a product. They're asking Google to compare options and give a recommendation.
"Which protein powder is best for muscle gain without bloating?" "Which CRM handles multi-currency billing without per-seat pricing?" These are decision support queries. The user is close to buying but specifically looking for comparative judgment — not a brand's self-description.
Your product ad copy is written to win on your own terms. AI Mode users asking "which" questions want an answer that situates your product against alternatives. Google's AI generates that comparative answer. Your ads appear in and around it. If your headlines and descriptions don't participate in the comparison frame, you're present for the query but absent from the conversation that's actually happening.
Writing ads that acknowledge the comparison context ("Why [product] outperforms on X" or "For people who've already tried Y") performs differently than pure brand copy on this surface — and the distinction matters when "which" is your fastest-growing query type.
Multimodal is past the 15% threshold
One in six AI Mode searches is now non-text. Image-based search in AI Mode has grown over 40% month-over-month since launch — faster than any other input type on the surface.
For e-commerce brands, this changes what product discoverability means. A user who photographs a competitor's product and searches for similar options is not a keyword searcher. They're a visual searcher, and the variable that determines whether your product appears is image quality and Merchant Center metadata, not bid strategy.
Shopping feed image requirements — accurate white or transparent backgrounds, multiple angles, lifestyle shots in the relevant supported categories — are becoming a targeting variable, not a creative preference. If your product images were set up when PMax launched and haven't been touched since, that's worth checking before those images determine your AI Mode eligibility.
Ads are growing on this surface
The zero-click rate on AI Mode is approximately 93%. Most intent resolves inside Google before a user visits any website. An earlier article on this data focused on that number and what it means for organic traffic.
Here's what's happening on the paid side: ads now appear in 25.5% of AI Mode results — up 394% year-over-year. Google's advertising revenue grew nearly 16% in its last reported quarter to $77 billion. The clicks-to-site metric is being replaced by something closer to share-of-answer — whether your brand appears in the AI-generated response, in the ad formats that surface alongside it, or in neither.
The formats built for AI Mode — Conversational Discovery Ads and Highlighted Answers — use natural language copy and product context. If you're running Performance Max or AI Max for Search, your existing asset quality score is already the admission ticket. Ad copy written for keyword matching will score lower, serve less, and reach fewer of the 1 billion users now on this surface.
What to do with this data
Your RSA asset library needs longer headline options. Not because longer is inherently better, but because AI Mode matches ad assets to conversational queries. A 6-word headline has almost no surface area to connect with a 35-word natural language query. The system tries, but it's working against the mismatch.
Write some headlines as questions: "Looking for a running shoe designed for flat feet?" Write some as comparisons: "How [product] handles [specific use case] vs. the alternatives." Those syntactic patterns match how AI Mode users actually phrase their queries. Your ads need to speak the same grammar.
Pull your Performance Max asset quality scores broken down by format. Video and image assets rated below "Good" now affect AI Mode placement eligibility. The images you submitted when the campaign launched may not qualify for the surface you're trying to reach.
If your campaign structure was built before AI Mode had a billion users, the Gromerce free audit surfaces where the gaps are between your current setup and what this surface actually rewards.
The keyword era isn't over. It's now sharing the budget with a surface that runs on full sentences.
Sources: Search Engine Journal, CMSWire, PPC.land, Digital Applied, SEOVendor, June 2026
