Google Ads built a two-decade business on a simple premise: you know exactly what keyword triggered your ad. You bid on "running shoes," someone searches "running shoes," your ad appears. The intent was explicit. The connection was transparent. You could optimize based on what you knew.
AI Mode breaks that contract.
Google's AI Mode, now live in the US with Shopping features rolling out in 2026, doesn't serve ads based on keyword triggers. It infers intent from conversational context across an entire session. The user didn't type a keyword. They had a conversation. And Google decided your product was relevant based on reasoning you can't inspect.
What Transparent Bidding Actually Was
Traditional search advertising was remarkably legible. A user typed a query. Google matched that query against your keywords. If the match qualified, your ad entered an auction. The auction had rules. The result was traceable.
Every dollar you spent could be connected to a specific search term. Not always the one you expected, but available in your search term report. You could see the exact query that triggered an impression, a click, a conversion. You could exclude terms that wasted money. You could double down on terms that converted.
This transparency wasn't a courtesy. It was the core product. Advertisers paid more over time because the data feedback loop made their campaigns smarter. You learned what your customers actually searched for, and you used that to build better landing pages, better products, better targeting.
AI Mode removes that feedback loop at the point of intent.
The Inference Black Box
In AI Mode, a user might ask: "I'm training for my first marathon and my knees have been bothering me. What should I look for in a shoe?" That's not a keyword. That's a statement with embedded intent, physical context, and purchasing motivation all mixed together.
Google's model infers from that conversation that the user is a beginner runner with joint sensitivity who is likely to buy within a short decision window. It surfaces a Shopping ad for cushioned stability trainers. The advertiser's ad shows up. The advertiser has no record of what query triggered the placement.
The intent was real. The conversion could be high. But the advertiser can't see the signal that drove it.
"You're paying for a placement you can't trace back to a trigger. That's a fundamentally different business than keyword advertising."
This matters for three reasons. First, you can't exclude irrelevant intent patterns the way you could exclude irrelevant keywords. Second, your quality score optimization assumes visible query data. Third, your attribution models are built on keyword-to-conversion chains that no longer exist in this format.
What Advertisers Are Actually Losing
The transparency loss is compounding. When you could see search terms, you could:
Build negative keyword lists that prevented wasted spend on low-intent queries. That capability doesn't translate to inference-based intent. You can't say "exclude users whose AI session included the word 'cheap'" because you don't have access to the conversation context.
Optimize landing pages to match the specific language users were using. If your search term report showed that people were finding you via "knee-friendly marathon shoe," you could use that exact phrase on your landing page. That feedback loop is gone.
Understand your real audience. Search term reports were the most honest market research available. What people type into Google when they're ready to buy tells you more about your customer than any survey. AI Mode conversations aren't shared with advertisers.
The Adaptation That's Actually Available
This isn't an argument to stop advertising in AI Mode. The reach is real and the conversion intent in conversational queries is high. The argument is to adapt how you measure and optimize.
Shift primary optimization to outcomes, not triggers. If you can't see what triggered the placement, optimize harder on what happens after the click. Conversion rate on landing pages, time to purchase, cart value. The signal moved downstream.
Invest in first-party data to close the attribution gap. The inference-based intent that Google is using is essentially matching your product to a user profile. The more your first-party CRM data matches your campaign targeting, the better your placements perform even without keyword-level transparency.
Test creative against audience segments, not queries. Creative optimization in AI Mode needs to work across ambiguous intents. Test broadly. A shoe ad that works for a cautious beginner and an experienced speed runner without knowing which one you're talking to is a different creative challenge than query-matched ads.
The shift from explicit to inferred intent isn't going away. Google's long-term direction is clear. Advertisers who treat this as a temporary disruption will keep trying to optimize for a transparency that no longer exists. The ones who adapt their measurement frameworks now will have a head start on the next version of performance marketing.
For the full picture on how AI is reshaping search behavior, the breakdown of Google AI Mode's effect on organic search runs alongside this one.