PPC After the Agentic Turn: A Q3 2026 Forecast for Marketers Who Still Have to Explain Why

PPC After the Agentic Turn: Q3 2026 Forecast

A Q3 2026 forecast for performance marketers and agency owners - what Google I/O and Google Marketing Live 2026 actually change for paid search: surface compression, AI Max automation, agent-mediated commerce, and the widening gap between what happens in your account and what you can explain.

---

A note on what is fact and what is forecast

Everything Google announced at Google I/O on May 19 and Google Marketing Live on May 20, 2026 is factual and sourced - the products, the protocols, the rollout windows. The percentage bands, scenario weights, and Q3 trajectories in this piece are Prooflytics analysis: directional scenario inputs for your own planning, not published numbers. Cited product launches are real. Forward projections are not. Re-baseline as the quarter unfolds.

---

Key takeaways

Google AI Mode shifts paid inventory from search pages into synthesized answers

Ads are migrating into AI-generated responses, not just declining in volume on traditional search pages. Planning for this environment requires treating the entry point as moving, not just contracting.

Informational queries lose paid value fastest as AI answers them in place

Top-of-funnel and comparison queries are the hardest hit by AI answer engines, while branded and transactional intents hold their paid value. Re-weighting budgets toward bottom-of-funnel is the structural adjustment most performance teams have not yet made.

AI Max replaced Dynamic Search Ads for all Google advertisers in 2026

Ask Advisor now runs diagnostics, builds campaigns, and writes reports across Ads, Analytics, and Merchant Center autonomously. Performance may improve; visibility into why it changed does not.

Agentic commerce platforms allow AI to assemble and checkout multi-merchant carts

Universal Cart, the Universal Commerce Protocol, and Agent Payments Protocol enable an AI agent to build a cross-merchant cart and check out without human input. A brand's job shifts from winning the click to being machine-readable enough to be selected by an agent.

Explaining why numbers moved becomes the scarcest skill in an automated marketing stack

When Google automates bidding, targeting, creative, and reporting, the one capability the platform cannot give back is a defensible answer to why a metric changed. Protecting this explanatory capability is the critical work for marketing teams in Q3 2026.

01 - The GML 2026 line

Where PPC actually stood the week everything was announced

The baseline matters because every Q3 projection is a delta against it, and the baseline at this moment is unusually legible because Google spent two consecutive days describing it out loud.

At Google I/O on May 19, the company reframed Search itself around an AI-powered "intelligent search box," generative UI that builds custom interactive layouts per query, and "information agents" that monitor the web on a user's behalf. Google reported AI Overviews at more than 2.5 billion monthly users and its conversational AI Mode at over 1 billion monthly users - these are not pilots, they are the default surface for a large share of search behavior. Generative UI rolls out free this summer; information agents and mini-apps built on the Antigravity platform land first for AI Pro ($19.99/mo) and AI Ultra ($99.99/mo) subscribers.

One day later, at Google Marketing Live on May 20, the advertising stack caught up to the search stack. Google's framing was explicit: the company said it is "moving from marketing automation to marketing intelligence," toward "a world where AI handles the complexity of execution and allows you to focus on what matters most." Read that sentence twice. It is the entire forecast in one line - and the entire question this piece exists to ask.

The concrete announcements that reset the baseline:

• AI Max - already the replacement for Dynamic Search Ads across all advertisers since April 2026, now extended to Shopping campaigns.
• Ask Advisor - a single Gemini agent spanning Google Ads, Analytics, Merchant Center, and the Marketing Platform, capable of running diagnostics, spinning up campaigns to target a new segment, and generating on-demand performance reports without the marketer touching each tool.
• Universal Cart + UCP + AP2 - a cross-merchant cart and the protocols beneath it, letting agents transact directly with merchant systems.
• Conversational ad formats inside AI Mode and Search, plus a new Qualified Future Conversions metric that ties brand-search actions to projected future sales.

The baseline shape

Automation was already dominant before this week - Smart Bidding, Performance Max, broad match, AI Max. What changed at GML 2026 is that the search surface and the execution layer are now being rebuilt in parallel, both around Gemini, and both in a direction that trades advertiser visibility for platform-managed performance. The magnitude is not the story. The composition is.

---

02 - The search-surface shift

Paid entry points move into the answer

The single most common mistake in reading this shift is treating it as "fewer impressions." It is not primarily a volume story. It is a placement story: the unit of paid inventory is moving from a ranked list of links to a slot inside a synthesized answer or a generative-UI widget that Google assembles on the fly.

That matters because the line between "answer" and "ad" - obvious in the ten-blue-links era - becomes deliberately blurry inside an AI response. Google is testing conversational ad formats directly in AI Mode. For the platform, this is more valuable than a clearly-labeled link; for the advertiser, it means less control over context, format, and adjacency.

The decline is uneven by intent - exactly as the organic shift has been.

Projected paid-entry-point health by intent · Q3 2026

The actionable read: a paid program weighted toward broad informational and generic mid-funnel terms is exposed on the surface that is being absorbed first. A program weighted toward commercial and branded intent is far better insulated. Re-weighting toward the bottom of the funnel is the highest-leverage move of the next ninety days - and the move most teams have not yet committed budget to.

---

03 - AI Max and the automation black box

Performance up, visibility down - and that trade is now structural

This is the section that matters most for anyone who has to answer to a client, a CMO, or a board.

AI Max replaced Dynamic Search Ads for everyone. Ask Advisor now sits in the Google Ads overview as a side panel that, in Google's own demo, surfaced a 15% sales lift, attributed it to updated product descriptions, and cross-referenced Analytics and Merchant Center data - all inside a single agent response. Google describes the back end plainly: "our agents talk to one another and carry each other's content, creating a continuous thread of intelligence."

That is genuinely useful. It is also a profound shift in who holds the explanation.

In the old stack, when CPL rose on a Tuesday, a competent marketer could open the account and reconstruct the chain: this match type expanded, that competitor entered the auction, this landing page slowed. The explanation lived in the data, and the marketer assembled it. In the AI Max plus Ask Advisor stack, the bidding, the targeting expansion, the asset combinations, and increasingly the narrative about what happened are all produced inside the platform. The agent hands you a conclusion. It does not hand you the reasoning you can independently verify.

The cost of this trade is not performance. Performance may well improve. The cost is defensibility - the ability to stand in front of someone who is paying you and explain, with confidence you can trace, why the number moved. When the explanation itself is generated by the same system that made the decision, the marketer has outsourced not just the work but the account of the work.

The cleanest illustration is the metric section 01 flagged in passing: Qualified Future Conversions. Powered by Gemini, it sits in the campaign view beside ordinary conversions and predicts sales up to six months out from signals like follow-up branded searches. The operative word is predicted. An ordinary conversion is a fact - someone bought, the tag fired, it can be checked. A Qualified Future Conversion is a forecast the platform's own model produces about a purchase that has not happened, scored against criteria the platform itself defines as "qualified." The advertiser ends up optimizing budget toward a number Google generated, graded by standards Google set, for an outcome no one can yet verify. The problem it addresses is real - AI surfaces absorbed the top of the funnel, so upper-funnel influence is genuine while the conversion is invisible - but the fix hands the scorekeeping to the same party selling the inventory. Branded search is the tell: brand demand rises for many reasons, and a forward-looking metric owned by the ad platform will tend to route that credit toward the upper-funnel spend it wants advertisers to keep buying. It belongs on the dashboard as a directional signal, not a KPI to fund against - and only beside an independent read of real brand demand.

This is the quiet structural risk of the agentic turn. Not "AI takes my job." Rather: "AI takes the part of my job that let me prove I was right."

---

04 - Bid dynamics on a shrinking commercial surface

The response is a re-mix, not a budget increase

The mechanical implication of a contracting and fragmenting paid surface is more spend chasing fewer high-intent slots. Base case: average CPCs rise 10-25% on competitive commercial intents through Q3, steepest in categories that lost the most informational and comparison surface - retail, SaaS, finance, travel. Part auction dynamics, part platform-side floor behavior responding to demand.

The reflexive response - raise the paid budget - is usually the wrong one in isolation. Budget poured onto the same broad commercial terms gets absorbed into higher bids with no ROAS uplift. The move is to re-mix where exposure sits.

Hold - Branded search

Branded CPCs stay relatively stable; the auction is defensive. Hold budget, hold bid strategy, hold share-of-voice. With AI surfaces now answering branded queries in-place, watch for branded-click leakage - but pulling back has a direct revenue cost. Defend the floor.

Re-mix - Generic commercial search

Generic commercial CPCs rise fastest. Shift toward higher-intent long-tail commercial terms where auctions are shallower; trim broad generic exposure where the rising floor erodes ROAS. Manage by intent depth, not headline spend.

Lean in - Mid-funnel paid social

Paid social mid-funnel surfaces hold cost stability better than search commercial in most categories, because the auction is less correlated with search-surface displacement. Lean into paid social for demand generation where search bids are rising disproportionately.

Test - Conversational and agent ad surfaces

Google is shipping conversational ad formats in AI Mode and is building ad surfaces into agentic shopping flows. Early entrants get low CPCs and uncrowded auctions. Allocate a 5-10% test budget to whichever surface lands first in your category - and instrument it so you can measure it, not just spend on it.

The teams that simply raised budgets in past cost-rise cycles saw the increase vanish into higher bids. The teams that re-mixed toward long-tail commercial, mid-funnel social, and an instrumented early test held ROAS within a few points of plan. The re-mix is the move; the spend question is downstream of it.

---

Prooflytics

Control paid performance across every channel

Every signal in one place. The whole picture. Your decision.

Start free trialSee pricing

14 days free · no credit card

05 - Agent-mediated commerce

The buyer is now intermediated. The brand surface changes shape.

Universal Cart lets a shopper assemble a single cart across Sephora, Target, Nike, Walmart, Wayfair, and Shopify merchants - from inside the Gemini app, from Search, from a YouTube video - and check out together via Google Pay. It runs on the Universal Commerce Protocol and the Agent Payments Protocol, which let Google's systems talk directly to merchant systems on the back end.

Small in absolute share today. Consequential in trajectory, because an agent standing between the buyer and the brand changes the merchandising surface in a way no prior channel has. The questions shift from "did they click my ad" to "did the agent surface my product, with the right attributes, at the moment of comparison."

Google is explicit that structured product data is the entry ticket. New Merchant Center tools - AI Performance Insights, Conversational Attributes - exist to make feeds machine-readable for conversational discovery, and to give early visibility into how brands perform inside AI shopping experiences. Watch how Google ends up defining "share of voice" across these surfaces; that definition will quietly set the rules of the next auction.

The readiness window

Agent commerce is a minority of purchases now and growing fast enough that the readiness work has to be done before the share is material: structured Product schema, complete attribute coverage, transparent pricing where public, frictionless checkout, comparison-ready content that cites primary sources. Programs that treated schema as a checkbox are exposed. The next two quarters are the window; after that it stops being an edge and becomes table stakes.

---

06 - The strategy response

What you keep when the platform automates everything else

Here is the turn.

Google's own thesis, stated at GML 2026, is that "AI handles the complexity of execution and allows you to focus on what matters most." The market has largely accepted the first half of that sentence. The unexamined part is the second: focus on what, exactly?

If the agent bids, targets, builds, and reports - and the report is generated by the same agent that made the decisions - then "what matters most" cannot be execution, because execution has been removed from your hands. It also cannot be the platform's own explanation, because that explanation is not independently verifiable. What is left, and what becomes scarce precisely because everything around it was automated, is the ability to understand why - across every signal, not just the ones one platform chooses to show you, and to decide for yourself whether the agent's conclusion is the right one.

Note the language Google used for the whole shift: "from marketing automation to marketing intelligence." That is the correct category. The open question is whose intelligence - the platform's, served back to you as a conclusion, or yours, built from sources the platform does not control.

The four things worth protecting in Q3, when execution is no longer the differentiator:

• Cross-source causal explanation. A single platform's agent sees a single platform's data. The marketer's edge is synthesizing what AI Max did with what changed at competitors, what shifted in the market, and what the platform updated - into one account of why that no single agent produces.
• An independent measurement layer. A monthly re-baseline of what actually happened in your accounts, held outside the platform's own narrative, so you can check the agent rather than take its word.
• The bottom of the funnel. Branded and transactional intent is where paid value concentrates as everything above it is absorbed. Defend it deliberately.
• The explanation you can hand a human. A client, a CMO, a board does not want the agent's dashboard. They want a person who can stand behind why the number moved. That is the deliverable automation does not produce.

The discipline behind all four is the same: every signal, the whole picture, your decision. Automation can run the campaign. It cannot, on its own, take you from a number that moved to a reason you can prove. That last step - from insight to proof - is the work that survives the agentic turn, and the work to build the next two quarters around.

---

07 - Scenarios and watch list

Eight Q3 scenarios, ranked by subjective weight at the GML 2026 line

Weights are subjective probability, not statistical inference. Re-weight as signals land. They are interior to each row, not normalized across the table - read them as independent watch items.

1. Base case - paid surface fragments, CPCs +10-25%, informational paid value compresses. Continuation of the trajectory; primary plan. ~55%
2. Conversational ads in AI Mode go broad fast - format becomes a real allocation question by Q3. ~30%
3. Aggressive case - AI Max defaults tighten further, manual control narrows. ~25%
4. Universal Cart adoption overshoots in retail - agent commerce gets material early. ~18%
5. Branded-click leakage - AI surfaces answer branded queries in-place, denting defensive paid. ~20%
6. Measurement opacity backlash - advertisers push for, and get, more PMax/AI Max transparency reporting. ~22%
7. Regulator pressure (EU/US) slows AI-search ad integration. ~12%
8. Macro ad-spend pullback compresses auctions broadly. ~12%

Watch list - signals that should trigger an off-cycle re-forecast

• Signal A - Conversational ad formats reach general availability in AI Mode. The test-budget line becomes a real allocation decision on launch day. Same-week re-forecast. Owner: paid-media lead.
• Signal B - AI Max forced-migration defaults change. Any tightening of defaults or narrowing of manual controls shifts the visibility-versus-performance trade. Same-week re-forecast. Owner: paid-media lead.
• Signal C - Universal Cart / UCP merchant adoption steps up in your category. Agent-commerce readiness moves from "next quarter" to "now." Same-week re-forecast. Owner: ecommerce / feed lead.
• Signal D - Google publishes how it defines "share of voice" in AI shopping. That definition sets the next auction's rules. Same-week re-forecast. Owner: strategy lead.

Four signals, four named owners, reviewed at the start of every monthly re-baseline. Short on purpose - a watch list that tracks twenty signals gets ignored.

---

Conclusion

The agentic turn does not retire the marketer. It retires the marketer who could only execute.

Google spent two days describing a world where the search surface and the ad stack are both rebuilt around an AI that handles execution end to end. The honest read is that a great deal of the manual craft of PPC - the bidding, the match-type management, the asset rotation, even the first draft of the performance report - is moving inside the platform and is not coming back.

What does not move inside the platform is the answer to why. When the same system makes the decision and writes the explanation, the explanation is no longer independent - and an explanation you cannot verify is not an explanation a client, a CMO, or a board can act on. The scarce capability for the next two quarters is the one Google's own framing pointed straight at: marketing intelligence - understanding the whole picture across sources the platform does not control, and deciding for yourself.

Re-weight toward the bottom of the funnel. Re-mix paid rather than just raising it. Harden agent-commerce readiness now, while it is still an edge. And hold an independent line of sight into your own accounts, so that when the agent hands you a conclusion, you are the one who can check it. That is the difference between running campaigns and being able to prove they worked - and in the agentic era, only one of those is defensible.

Prooflytics surfaces this cross-source picture in your daily briefing - combining platform signals, competitive moves, and your own account data into one independent view, so the explanation is yours to verify.

---

Outside Google's ecosystem, OpenAI launched a self-serve ChatGPT Ads platform in May 2026 — $3-5 CPC, no minimum spend, conversational placement targeting. For performance marketers deciding whether to allocate test budget to ChatGPT alongside their Google campaigns, the ChatGPT Ads channel reality check covers what the attribution data shows and when the channel makes sense in a multi-channel mix.

Frequently asked questions

How is the "paid surface compression" measured, and how confident is the projection?

The compression here is qualitative-first: the observable shift is that paid inventory is moving from ranked links into AI-answer slots and generative-UI widgets, confirmed by Google's own launch of conversational ad formats in AI Mode at GML 2026. The CPC band (+10-25% on competitive commercial intents) is a Prooflytics base-case estimate, not a published figure - re-baseline monthly against your own account data.

Does this mean Google Search PPC is dying?

No. It means the high-value paid surface is concentrating at the bottom of the funnel - branded and transactional intent - while informational and broad mid-funnel paid value erodes as AI answers those queries in-place. A program built on bottom-funnel commercial intent is well-positioned; one built on broad informational traffic is exposed.

What is the difference between AI Max and Ask Advisor?

AI Max is the automated campaign engine that replaced Dynamic Search Ads - it handles matching, targeting expansion, and asset combination. Ask Advisor is a Gemini agent layered across Google Ads, Analytics, and Merchant Center that runs diagnostics, builds campaigns, and generates reports. Together they automate both the execution and the narrative about the execution.

Why is "explainability" the thing to protect rather than performance?

Because performance is what the platform now optimizes for you, while explanation is what it does not give back in a verifiable form. The marketer's defensible value to a paying client or an executive is a traceable answer to why results changed - and when that answer is generated by the same agent that made the decisions, it stops being independent. Protecting an independent line of sight is protecting your defensibility.

What should a mid-sized B2B SaaS or agency team actually do in the next ninety days?

Four things in parallel: re-weight paid toward bottom-funnel commercial and branded intent; re-mix rather than inflate spend (long-tail commercial, mid-funnel social, a small instrumented test on conversational ad surfaces); start agent-commerce readiness if you sell products (structured feeds, schema, comparison-ready content); and stand up a monthly re-baseline that holds your account truth outside the platform's own reporting.

Is agent-mediated commerce relevant if I sell services, not products?

Less directly today, but the underlying shift applies. Information agents and AI Mode increasingly mediate high-consideration research - exactly the journey B2B and service buyers take. The readiness work shifts from product feeds to comparison-ready, citation-quality content that an agent can surface and trust. The window to invest is the same two quarters.

---

This forecast is directional analysis for planning, not deterministic prediction. Re-baseline at the end of each month; let the watch list trigger off-cycle re-forecasts when a signal lands.