Keywords Are Dead: Operators Must Master Intent Signals

TL;DR: Google’s keyword system is being replaced by intent-based matching, and the data proves it: Exact Match’s share of non-branded spend collapsed from 37.1% in 2022 to 27.6% today. For operators in high-CAC verticals, the shift from keyword lists to business signals is not optional. Your landing pages, feed quality, and offline conversion data are now your targeting inputs.

The Deal That Held for Two Decades

For most of Google Ads’ history, keywords were a contract. You researched them, structured your account around them, and wrote ads that matched their intent. Google, in return, promised your ads would only surface when those specific queries fired in the auction. Exact meant exact. Phrase meant phrase. Broad was the reach lever you pulled carefully, if at all.

That arrangement gave operators something worth protecting: diagnosability. A campaign underperformed, you opened the search terms report, identified bad queries, added negatives, and promoted winners. The process was repeatable, teachable, and — for a long time — reliable. The author of this analysis, who joined Google in 2002 as one of its first AdWords evangelists, describes this as the system he helped build. It worked because the underlying search technology had real limits, and keyword lists filled the gap.

That gap has closed. The contract is void.

How the Shift Happened, One Product Decision at a Time

The breakdown was gradual. Close variants first expanded Exact Match to cover misspellings and plurals, then function-word variations. By the mid-2010s, “exact match” was already a soft approximation. Smart Bidding shifted auction logic from keyword triggers to conversion probability — which user converts, not which keyword fired. The 2023 Broad Match rebuild made Broad semantically competitive, reporting roughly 25% more conversions in Target CPA campaigns and flipping the narrative for advertisers who had spent a decade treating Broad as a budget drain.

AI Max is the culmination. Feed Google your URL, your creative assets, and your business data — and keywords become optional. The system generates “synthetic keywords” on the fly, matching your ad to the shape of a user’s conversational prompt rather than a phrase you pre-declared. OpenAI’s ad surface, where operators can now place ads on ChatGPT, also treats keywords as optional from day one. When the company that invented keyword advertising and the company reinventing search both land on intent-matching with no required keyword input, the direction is obvious.

Operators running paid search campaigns at scale need to understand this is not a feature update. It is an architectural replacement.

What the 2026 Match Type Data Actually Shows

The author’s firm analyzed nearly 130,000 non-branded campaigns across more than 14,000 accounts, totaling roughly $99 million in spend. The numbers are concrete:

• Exact Match non-branded spend share: down from 37.1% in 2022 to 27.6% today, with most of the drop in the last 24 months.
• Phrase Match now drives 40% of non-branded conversions at a 15.7% conversion rate — well above Exact’s 10.5% and Broad’s 8.5%. Phrase has become what Exact used to be: the default for scalable, intent-respectful discovery.
• Broad Match represents 38.8% of non-branded spend, the single largest bucket by volume.
• On branded terms, the calculus flips entirely. Exact Match delivers 6.61x ROAS at a $0.90 CPC, nearly double the ROAS of either alternative.

The practical playbook that emerges from practitioner data: launch new lead gen campaigns on manual CPC and Phrase Match. Collect clean traffic for several weeks. Promote proven winners to Exact. Layer Broad and Smart Bidding on top only after the algorithm has something real to optimize against. Exact is “too specific for a new campaign.” Broad is “overly aggressive at the start.” Phrase is the middle position — flexible enough to surface intent you didn’t anticipate, tight enough to keep data actionable.

What Operators Are Losing

This transition is not cost-free. Three things degrade as keyword control loosens.

Granular diagnosability. When an intent-matched campaign underperforms, the old debugging sequence — search terms report, add negatives, tighten match type — only partially applies. Negative keywords still function and still matter, but the matching engine is harder to reason about. “Why did my ad show there?” has a fuzzier answer in 2026 than it did in 2016.

Account structure as craft. Tight ad groups, themed structures, clean branded-versus-non-branded separation — much of this was a proxy for control. Once the system handles more of the targeting, elaborate structure loses strategic value. Some of it was over-engineering. But the discipline of mapping intent to campaigns carefully built real judgment, and that skill risks atrophying when the system absorbs it.

Training signal for junior operators. The search terms report was a masterclass in consumer psychology. You watched query patterns for weeks and built intuition about how users phrase problems, how language shifts seasonally, how intent clusters. A system that abstracts the keyword also abstracts away one of the best diagnostic tools the discipline has ever produced — and removes early visibility into shifts in consumer behavior that would otherwise inform strategy.

Operators who rely on precise audience targeting at the query level will feel this the most acutely in the short term.

What High-CAC Operators Are Gaining

For operators in verticals where a single conversion can be worth thousands of dollars — forex, iGaming, legal, crypto — the gains are meaningful if you feed the system correctly.

Intent-based matching covers queries no keyword list ever catches: zero-click conversational prompts, brand-new phrases, generational slang, hyper-local variations. These are exactly the places manual selection fails — not because operators are inattentive, but because the query space is too large and too dynamic to enumerate. iGaming acquisition teams and forex lead generation programs running on tight keyword lists are already leaving coverage on the table.

The maintenance overhead drops substantially. Negative keyword lists stretching into the thousands, endless query audits, SKAG construction, quarterly match type experiments — much of that work was overhead imposed by the gap between advertiser language and user language. Closing that gap algorithmically frees up hours for strategy, creative, and measurement. That reallocation compounds over time.

Access to richer signals also improves. Google’s LLM-driven query understanding sees more of a user’s journey than any keyword list will. Operators unwilling to let those signals into their targeting are competing with less information than their opponents. That is a structural disadvantage in a high-stakes auction.

What This Means for High-CAC Vertical Operators

Operators in legal, crypto, iGaming, and forex are already paying some of the highest CPCs in Google’s auction. The margin for structural errors is thin. Here is what this transition demands specifically:

Separate branded and non-branded campaigns — hard. The data shows 6.61x ROAS on branded Exact versus roughly 3x when mixed. Sloppy brand separation is a direct tax on return. Lock branded to Exact, protect it, and do not let non-branded bleed in. Law firm marketing programs and crypto acquisition campaigns that have not done this separation are paying for it in ROAS degradation right now.

Treat your business signals as your new targeting infrastructure. Landing page quality, asset library hygiene, product feed accuracy, and — most critically — offline conversion data piped back to Google. The qualification gap between “cost per lead” and “cost per qualified lead” is often the difference between a campaign that looks good on a dashboard and one that actually generates revenue. If you’re running 100+ leads per month, your CRM integration with Smart Bidding is not optional. It is the primary lever.

Maintain negative keywords aggressively. In the AI Max environment, negatives are the last hard boundary you control. Brand safety, budget discipline, and irrelevance prevention now live here. Automate additions where possible. This is not where you want manual lag.

Upgrade the skill stack. The future PPC operator is not a keyword picker — they are an intent engineer who can translate business goals into signals the algorithm learns from, debug a semi-opaque system using query reports and holdout experiments, and explain account performance to clients when the dashboard only shows aggregates. A structured paid search audit against your current keyword architecture is the fastest way to identify where legacy structure is costing you efficiency in the new system.

For teams considering AI-assisted lead qualification on top of intent-driven traffic, AI agents for lead qualification can absorb inbound volume from broader match targeting without degrading close rates — a practical pairing as match types loosen and query volume expands.

The Transition Is Already Halfway Done

The industry has been moving toward intent-driven matching for years. AI Max accelerates the final leg of that journey. The keyword as an advertiser artifact — a pre-declared list you maintain and optimize — is being replaced by a system where your business signals, your page quality, and your conversion data are the targeting inputs. Operators who adapt early gain the efficiency advantage. Those who defend legacy keyword structures will find themselves competing with less information, slower coverage, and higher effective CPAs. The question is not whether to adapt. It is how fast.

Originally reported by Search Engine Journal, May 2026.