AI Overviews reshaped the results page, yet many paid search playbooks still follow the same rule: bid up where performance looks strong, bid down where it looks weak. That assumes every query is the same kind of auction. Ana Kostic, co-founder of Spanish performance agency bigmomo, argues the assumption breaks in B2B and industrial accounts with long sales cycles and thin conversion data. When Google answers directly, the value of a paid click varies by query — so an AI Overviews paid search strategy starts with classification, not bids.
Bottom-funnel queries with clear commercial intent where a click still becomes a lead or sale: product plus modifier, supplier shortlists, “buy,” “quote,” “distributor,” brand terms with buying intent. A July SE Ranking study of commercial queries in U.S. AI Mode found ads appeared about 54% of the time for keywords costing $10 or more per click, 32% between $2 and $10, and around 24% under $2 — and about seven in 10 answers with ads showed two advertisers. Pay the premium only where absolute top still beats your incremental CPA, estimate costs with bid simulators, and watch Absolute Top impression share and impression share lost to rank.
Queries where the AI answer does the educating and users often don’t click. The goal is to be cited or named in the answer, not to buy a cheap visit. As Google’s Shashi Thakur said at Google Marketing Live EMEA 2026, the best ads must be answers. In one of Kostic’s B2B accounts, terms triggering AI answers were 83.3% informational, 10.0% learn-and-solve and 6.7% commercial. Plan for share of the answer rather than maximum CPC, coordinate with SEO and avoid paying where organic already wins.
Queries where zero-click has already won. You can keep bidding as CPCs climb and volume falls, or fund work that happens before the search: Demand Gen, YouTube, first-party content, reviews and community presence that pay off later when the same person searches with intent. Measured on last-click ROAS it always looks weak, so track brand search and pipeline.
Two foundations come first: semantic coherence — keyword, ad and landing page as one unit of meaning — and conversion quality, including lead scoring and offline or CRM signals. Then:
AI Max and broad match sit on top of this classification; they don’t replace it. They’re needed to appear around AI answers but reduce control, so introduce them through small tests where the bucket is clear and tracking is clean.
Accounts that apply one rule to every query keep funding the wrong mechanism — the same shift behind rising Shopping CTR alongside falling impressions and AI Overviews contradicting the ads above them. Moving from bid optimization to budget reallocation across buckets is how paid search adapts to how people search now.
Fight covers clear-intent queries worth paying for top position; Influence covers queries where AI answers educate and the goal is citation; Generate demand covers zero-click queries where budget should move to pre-search channels.
In an SE Ranking study, ads appeared on about 54% of commercial queries with CPCs of $10 or more versus about 24% under $2 — paid inventory follows commercial value.
No. They help ads appear around AI answers but reduce control, so use them through small tests where the bucket is clear and tracking is clean.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
When AI Max expanded into competitor searches an advertiser deliberately avoided, it proved automation needs checks. Mike Ryan on AI Max guardrails and campaign structure.
Paid search veteran Matt Van Wagner reflects on two decades of Google Ads, lost tactics and automation — and why advertisers must never assume Google acts in their interest.
Google is testing text ads from exact and phrase match keywords in AI Mode when intent is explicit. A possible path into AI Mode without AI Max or PMax.