Google Ads Split Into Three Businesses. Your Plan Should Too.

The Chart Nobody Put on One Slide

When Alphabet reported Q1 2026 in late April, the headline coverage was "Search +19%, beat estimates." That headline was correct and badly incomplete. The full segment picture from the disclosure looked like this:

Three lines, three directions. The blended +15.5% looks healthy. The dispersion underneath it is the actual story.

For an SMB allocating media budget across Search, YouTube and Demand Gen, and any Display or programmatic placement served through the Google Display Network, those three line items are no longer parts of the same machine. They are parts of three machines in three states: one accelerating, one accommodating, one contracting. Treating them as a single channel with a single strategy was already lossy in 2024. In 2026 it is the kind of planning error that quietly mis-prices an annual budget.

Search (+19%): The Strong Half Is Getting Stronger and More AI-Mediated

Search is the part of Google getting noticeably stronger. Google's own commentary from the earnings call: queries at all-time highs across major markets, AI Overviews now reaching about 2 billion monthly users across 200+ countries, and "people coming back to Search more" via AI Mode.

Statcounter still puts Google at roughly 90% of worldwide search engine share and 85% in the US, so the +19% on Search is happening from a position of dominance, not from a recovery base. It is also happening with cost relief rather than cost pressure: Tinuiti's data shows brand text CPC inflation that spiked to +19% YoY in Q1 2025 had eased to +2% by Q4 2025, and non-brand CPCs went mildly negative through Q3. Some of that easing has a non-Google explanation. Amazon withdrew from US Google Shopping auctions on 23 July 2025, removing a uniquely large bidder and pulling the Shopping CPC trend with it for the rest of the year. The Search & other +19% sits on top of a quieter, partly conditional cost picture, not a uniformly inflationary one.

What is underneath that growth is a Search product that does not look like the 2018 version. AI Overviews have carried their own ad inventory at scale since the May 2025 rollout, with eligibility now broad across English-language markets. AI Mode is a smaller, still-emerging surface: ads were in test for most of 2025 and only expanded to sponsored retailer formats and Direct Offers in February 2026, including for travel. By volume the AI inventory most SMB accounts will engage with in 2026 is AI Overviews, with AI Mode as a secondary layer that will matter more as 2026 progresses. The campaign types eligible to fill those AI placements are broad match, AI Max, Performance Max, Shopping, and DSA. Exact and phrase match keywords cannot trigger them.

Translation: the part of Google that is growing is the part advertisers have least manual control over. Search & other revenue grew 19% because Google wedged ads into AI surfaces and built campaign types that feed them. The "old Google Ads" account, the one a Western Canadian SMB used to run with exact-match keyword discipline and tight ad-group structure, does not have inventory access to the surfaces driving that +19%.

That sentence is the load-bearing observation in this post. The platform did not just add an AI feature; it shifted the inventory funding the headline growth into surfaces with a different control regime. Every other recommendation in this piece descends from that fact.

Aaron Levy, now Optmyzr's evangelist and previously VP of Paid Search at Tinuiti, describes what the underlying behaviour looks like across the Optmyzr account base: "click and traffic volume was down like 15 to 20% year-over-year. Revenue was up, conversion rate was up, conversion volume was up." His read is that AI is pre-qualifying users and Search is becoming "a true bottom-of-funnel catch-all." Levy is also direct on the cannibalization mechanism: PMax and AI Max, in his words, "heavily overindex on competitive queries... on one side you're paying higher CPCs to bid on a competitor's brand name, and on the other side you're getting your own brand CPCs inflated as well." That is the part SMBs can watch for in their own accounts. Brand-term CPC creep that does not match search-volume changes is a leading indicator that PMax or AI Max is bidding inside the brand auction, on both sides of it.

This is a fork in the road that does not get framed as a fork in the road. Either an account engages with broad match, AI Max, and Performance Max with proper guardrails and joins the inventory expansion, or the account remains in a tight exact-and-phrase posture and quietly cedes share of voice on every AI-eligible surface. Both paths are defensible. Defaulting into one of them by accident is not.

YouTube (+11%): Pricing Accommodation, Not Pricing Power

YouTube's headline +11% looks like the calm middle of the three. Tinuiti's data through Q4 2025 shows what is happening underneath: YouTube CPMs were down approximately 18% year-over-year even as impressions surged. Volume is up, unit price is meaningfully down, and aggregate revenue is growing because volume is growing faster than price is falling.

The mechanism is competitive. YouTube, Demand Gen, and Reels-class video compete for the same dollars as Meta Reels, TikTok, and connected-TV inventory. Meta's Advantage+ video automation has matured to the point that EMARKETER attributes a meaningful chunk of Meta's 2026 acceleration to it. To keep YouTube impression growth alive against that competitive set, Google has had to accommodate price.

For SMBs, that accommodation is a pricing window. Demand Gen at 2026 CPMs is structurally cheaper than Demand Gen at 2024 CPMs, and the conversion-tracking gap that used to plague YouTube has narrowed considerably with Google's enhanced conversion and Data Manager work. Whether you should fill that window depends on whether your business has a story video can actually tell: a real founder story, a treatment-results walkthrough, a service explainer that answers the questions sales gets every week. If you do, the discount is real, and there is no reason to assume it stays this generous through 2027.

Google Network (-4%): The Open-Web Ad Layer Is Contracting

This is the line that did not make most coverage. Google Network revenue (AdSense, AdMob, and Ad Manager partner inventory; the third-party publisher monetization layer) fell 4% YoY in Q1 2026.

That is the most pronounced Network decline in years, comparable in size only to the 2022 to 2023 ad-market downturn, and unlike that earlier dip it is happening while Google's owned-surface revenue is accelerating, not while the broader ad market is soft. The cleanest explanation is the one PPC Land surfaced bluntly: "Google Network ad revenue falls 4% as AI reshapes the web." When the queries that used to bounce out to publisher pages now resolve inside an AI Overview or an AI Mode panel, those publisher visits do not happen, those publisher impressions do not get sold, and Google's third-party network revenue contracts.

The AI assistant layer where those queries now resolve has fragmented at the same time. Statcounter's April 2026 chatbot referral data shows ChatGPT at 76.85% of global AI assistant referrals (down from 84.21% a year earlier), Gemini at 9.00% having overtaken Perplexity in March 2026, Perplexity at 7.73%, Microsoft Copilot at 3.76%, and Claude at 2.66% after growing nearly tenfold from April 2025. Statcounter CEO Aodhan Cullen's framing: "For website owners and digital marketers, the message is clear: optimizing for a single AI chatbot is no longer sufficient." The Network's -4% is the trailing indicator of that fragmentation; the leading indicator is what an AI Overview or a Gemini result does to a publisher's organic CTR before the impression ever reaches their ad stack.

For an SMB advertiser, the Network line is the inventory you reach when you run Display campaigns, GDN placements, programmatic buys executed through Google's stack, or Performance Max segments that hit display surfaces. That inventory is the contracting half of Google's ad business. The publishers carrying it have weaker traffic, less premium adjacency, and a worsening monetization model. Buying it as a flat replacement for Search reach was never quite right. In 2026 it is meaningfully wrong.

The line item that needs the hardest look on a current portfolio is Performance Max with all surfaces enabled. Without explicit campaign-level signals telling PMax to favour Search and YouTube, the algorithm is free to allocate spend toward Display and Network surfaces, which is exactly the inventory Google's own segment data is showing as the weakening half. PMax's appetite for cheap impressions is a feature in normal conditions and a problem when the cheap impressions are coming from a contracting publisher base.

Why the Three Businesses Are Diverging

The three numbers stack into one story: AI assistants are eating the open web, and Google's response is to defend its owned surfaces while the open layer shrinks.

Search and YouTube are owned. Google can ship ads into AI Overviews because Google operates AI Overviews. Google can compress YouTube CPMs to keep impression growth alive because Google operates YouTube. The Network is not owned. It depends on third-party publishers having traffic and being able to monetize it. Both halves of that loop are weakening: AI assistants intercept the queries before publishers see them, and the impressions that do reach publisher pages monetize less efficiently when premium content is being summarised upstream.

Google's product roadmap of the last twelve months reads off the same logic. AI Max for Search, ads in AI Overviews, ads in AI Mode, AI Max for Shopping and Travel, Direct Offers, agentic commerce, the Power Pack framing, the September 2026 forced migration of DSA and campaign-level broad match into AI Max: all of these push budget into Google-controlled, AI-mediated environments where Google handles matching, bidding, creative assembly, and inventory allocation. That is where the +19% on Search is coming from. The Network does not get a comparable defensive product because Google cannot ship a third-party publisher's traffic into existence.

The strategic implication is uncomfortable but clean. The more Google's growth depends on AI-mediated owned surfaces, the more advertisers are buying into a buying environment they have less direct control over. The trade is real reach for less granular governance, and the longer an SMB delays acknowledging it, the more the trade gets made by default rather than by decision.

Frederick Vallaeys, Optmyzr's CEO and a former Google AdWords Evangelist, frames the asymmetry directly: advertisers are often the ones "paying for Google to learn" as the automated systems mature. The R&D burden of building AI-mediated advertising is being borne in part by advertiser budgets through the learning curves of AI Max, Performance Max, and Smart Bidding Exploration. That is the cost of being early on a defensive product cycle, and it is the strongest argument for engaging with these systems deliberately, with measurement, rather than by accident.

The Meta Footnote That Is Actually a Headline

EMARKETER published a 2026 worldwide ad revenue projection in late Q1: Meta $243.46B at 26.8% share with +24.1% growth, and Google $239.54B at 26.4% share with +11.9% growth. This would be the first year Meta surpasses Google on total digital ad revenue.

EMARKETER attributes the gap to specific Meta levers: Advantage+ automation maturing into a default-on engine for many advertisers, AI-generated creative reducing production cost, Reels gaining attention share against YouTube and TikTok, and measurable advertiser ROI gains compounding from those three. Google is shipping a parallel automation playbook (AI Max, PMax, Demand Gen, Smart Bidding Exploration). Meta is currently executing the playbook on advertiser ROI more effectively.

For Western Canadian SMBs, the practical reading is this: the assumption that "Google is the default platform and Meta is supplementary" is an artifact of 2018 to 2022 buying behaviour. By the end of 2026 it may not be a defensible default. Treating Meta competency as a 20% retargeting line item, the way the Google Ads vs Meta Ads post historically scoped it for many professional services accounts, was right for 2024. By 2027 the same client's portfolio may need a more substantial Meta build-out, which means the agency and the client should be developing that capability now, not under deadline pressure when the Google numbers plateau.

What This Means for a Western Canadian SMB in 2026

The bifurcation is showing up in practitioner sentiment data, not just in revenue lines. The State of PPC Report 2026, surveying 1,306 PPC professionals across 50+ countries in late 2025, found 53% of respondents saying managing PPC is harder than two years ago, 48% citing "lack of control" as their top Performance Max frustration, and 62% identifying increasingly opaque, black-box platforms as their top overall industry challenge. SMBs are not imagining the new shape of the work.

Five practical implications.

1. Stop scoping "Google Ads" as one strategy. Search (with brand and non-brand sub-lines), AI Max, Performance Max, Demand Gen, and any Display exposure are now five different products operating in five different cost-and-control regimes. Budget conversations should treat them as separate line items with separate measurement plans, separate experimentation cadences, and separate accountability metrics. Blended "Google Ads ROAS" obscures more than it reveals; in 2026 it conceals the surfaces that matter most.

2. Don't bake Display or programmatic Google Network into 2026 to 2027 lead-gen plans without a measurement-validity check. The contracting half of Google's ad business is the half most SMBs are quietly using when they let Performance Max default to all surfaces or buy "Google Display reach" as a top-of-funnel layer. Where the inventory is shrinking, the placements available are skewing toward leftover inventory, not premium adjacency. Audit PMax channel performance reports campaign by campaign and decide explicitly whether Display and Network surfaces stay enabled.

3. The YouTube and Demand Gen pricing window is open right now. If you have video assets that can do real work, Demand Gen at 2026 CPMs is structurally cheaper than it has been for the last three years. Run a small structured test inside a dedicated campaign with its own conversion actions, not bundled into PMax. The discount is being subsidised by Google's competitive position; the position is unlikely to soften, which means the discount may not last.

4. Build real Meta competency in parallel. Not as a retargeting afterthought. As a primary-channel option for the right business profile. The clients who get burned in 2027 are the ones whose first serious Meta build-out happens after their Google performance plateaus and the budget has already been re-cut. Start the build now, with measured budgets and operational rigour, while the Google side is still funding the operating budget and there is room to experiment without it being a rescue mission.

5. Treat AI surfaces as their own line item with their own measurement plan. AI Overviews ads, AI Mode placements, and AI Max-driven query expansion are how Google's strong half stays strong. They also cannot be honestly measured by the same blended ROAS view that worked for keyword campaigns. Run them as separate campaigns with brand exclusions, holdouts where the geography allows, and a profit-margin lens rather than a topline conversion-volume lens. We covered the practitioner-side problem with AI Max attribution fog briefly in The Google Ads Job Changed in 2026; the point holds doubly when the inventory you are testing is part of the line of business Google is most aggressively defending.

What We Changed Across the Portfolio

When the Q1 2026 disclosure landed and the bifurcation became visible in the segment data, we made several concrete operational changes across our active accounts. The first was a reporting split: every account's monthly report now breaks Google spend into Search (with brand and non-brand sub-lines), AI Max, Performance Max, Demand Gen, and Display or Network exposure, with each surface getting its own conversion volume, cost per qualified lead, and margin column. Aggregating these into one "Google Ads" number was already losing information by 2024; in 2026 it loses the most important information. The four further changes:

Display and Network exposure audited and downgraded by default. Across the portfolio, Performance Max settings were reviewed for Display and Network surface inclusion. For lead-gen accounts, those surfaces were excluded by default unless a holdout test had proven incremental contribution. The contracting half of Google's ad business is not an automatic place to put SMB lead-gen budget, and PMax's defaulted surface mix was sending money there without explicit consent.

A YouTube and Demand Gen test slot opened in every account that has video assets. Small budgets, separate campaigns, separate conversion actions, separate reporting. The hypothesis is that current CPMs are competitive enough to warrant a controlled test, not a portfolio commitment. If the unit economics hold up over 60 to 90 days, the line item gets sized up. If they do not, it gets parked.

Meta build-out promoted from optional to recommended in 2026 plans. For new clients, the proposal now includes a Meta competency build phase even when the initial channel mix is Google-heavy. For existing clients in professional services, where the law firm case study historically pushed Meta down to a retargeting role, we re-scoped the Meta line as a 12-month capability investment rather than a fixed budget allocation, with a decision point at month nine on whether to size up the channel materially.

AI-surface measurement scaffolding added. AI Max and broad-match-eligible campaigns running into AI Overview and AI Mode placements now report through a separate measurement layer with explicit holdout tracking where geography supports it. This is the same operational discipline we built for tracking activity across 10 AI platforms on the SEO side; the paid media equivalent had lagged it, and we closed the gap this quarter.

The Short Version

The Google Ads platform you bought in 2020 was one product with a coherent strategy. The Google Ads platform you are buying in 2026 is three products moving in three different directions, with the strong half being absorbed into AI surfaces you have less direct control over and the weak half quietly contracting under your Display line items. Meta is closing on Google's total ad revenue while running a parallel automation playbook more effectively on advertiser ROI.

The right response is not panic and not denial. It is portfolio thinking. Treat Search, AI-led campaigns, video, and Display as four products, not one. Each gets its own measurement plan, budget envelope, and strategic question. Build Meta competency before you need it. Audit Display exposure as a contracting inventory layer rather than a flat reach option. Run AI surfaces as measured experiments, not defaults.

Read more about our Google Ads management approach, our Edmonton PPC practice, and our broader work on AI marketing. For the operational adjustments we made when the 2026 product changes started landing, see The Google Ads Job Changed in 2026 and our framework for where to spend your first paid-media dollar. To talk through what a portfolio-level rebuild would look like for your business, get in touch.