Q3 2026 Google Ads Roundup: What Worked, What Broke and What's Next
· PPC

Q3 rewarded the least glamorous work in the account.
Across our Google Ads book, the biggest wins came from keyword buildout, query routing, feed cleanup and Shopping structure, while the newer automation levers delivered mixed results that demanded close monitoring.
The auction also got more expensive. CPCs across our clients rose 25% QoQ, with the steepest pressure arriving after Google's August 17 bidding update. When every click costs more, a messy structure gets punished faster, which is why the fundamentals carried so much weight this quarter.
This roundup comes from Cindy, our Director of Paid Search, and the team running Google Ads across our ecommerce client book. We cover the restructures that drove the quarter's best results, where AI Max earned its place and where it burned budget, a branded search tactic that stopped working, and the one bidding test we would run before Q4 peaks.
The Basics Carried the Quarter
Every meaningful Q3 win in our book traced back to fundamentals. New client onboards showed it most clearly, because onboarding is where clean structure produces the fastest lift.
The work that won:
Thorough keyword buildout based on real query data
Clean query routing, so each search lands in the campaign built for it
Ongoing ad copy and landing page testing
Feed optimization
Shopping structures built on business data instead of default product groupings
Recurring account maintenance
Two restructures show what that looks like in practice.
Nonbrand Shopping restructure for an online pet supplies brand
We rebuilt Shopping around SKU quality. Margin and item class data from the business were combined with Google Ads revenue per click (RPC) to cap low-value tail products, which pushed more spend toward proven winners.
Conversion rate up 30%
New customer sales up 94%, from $43k to $84k
ROAS up 17%
Default Shopping structures let Google spread budget across the whole catalog. Tail SKUs absorb spend at low RPC, which starves the products that actually bring in new customers. Capping the tail gives the bid strategy a cleaner set of products to scale.
Build those caps from your own margin data. Google's reporting shows what converts, while your P&L shows what each conversion is worth to the business.
Nonbrand Search restructure for an oral pouch ecommerce brand
We started by cleaning up the existing structure. From there, we expanded keyword coverage based on query analysis, regrouped ad groups around stronger themes and refreshed RSA testing. Results are period over period (PoP) on a 4% increase in spend.
Revenue up 47%
ROAS up 42%
Conversions up 27%
CPA down 20%
CPC down 28%
A 47% revenue lift on nearly flat spend is what structural work looks like in the data. Tighter themes improve ad relevance, which pulls CPC down, while better routing sends each query to copy written for it. None of it required a new campaign type or a new feature.
AI Max Worked Where Intent Was Clean
AI Max for Search campaigns lets Google match beyond your keyword list and adapt ad copy to each query. In our book, it performed best in accounts with straightforward products, clean search intent and low regulatory complexity. Those accounts used it as an expansion lever for incremental traffic, although results still varied by campaign and query environment.
The problem-aware play
Our strongest AI Max result came from a structure Google does not recommend. We launched it in campaigns that reached beyond product and category terms into problem-aware searches, the queries people type when they know the pain point but have not picked a product yet.
Those queries drove strong efficiency. They also opened a new pool of nonbrand demand further up the funnel, which traditional category keywords were never going to reach.
If you test this, build the campaign around the problem your product solves. Review search terms every week to confirm AI Max is staying inside that intent.
Where it failed: sensitive verticals in the UK
AI Max underperformed for sensitive verticals in the UK, where serving on AI Overviews is not yet available. Of the queries AI Max matched, 80% were already covered by existing keywords. In practice, the account paid a 26% CPA premium on average for traffic its keyword set was already capturing.
Treat AI Max as a test in every account, never a default. A win in one account, or even one campaign, tells you nothing about the next one. Before you scale it, check query overlap against your existing keywords and compare CPA at the campaign level.
Manual CPC on Branded Search Stopped Working
For over a year, moving Branded Search from tROAS or tCPA to Manual CPC was one of our most reliable efficiency plays. Tests we ran last year and early this year delivered consistent efficiency gains as high as 60% while holding conversion volume steady.
That tactic no longer works in our accounts. Manual CPC is a deprioritized product that Google does not recommend. The effect of that deprioritization now shows up in our branded data.
If Manual CPC still runs on your brand campaigns, compare Q3 efficiency against your H1 baseline. Faded gains mean it is time to test alternatives, ideally before Q4 traffic makes every inefficiency more expensive. Our team is testing into the next best strategy now.
How Our Book Performed in Q3
Q3 spend across our Google Ads clients rose 37% QoQ, while a 25% jump in CPCs pushed efficiency in the wrong direction.
Metric | Q3 vs Q2 |
Spend | +37% |
CPC | +25% |
CPA | +17% |
ROAS | -10% |
Higher CPCs over time are expected, especially after Google's August 17 bidding update. The detail worth noticing is that CPA rose less than CPC did, which means conversion rate improved underneath the cost pressure.
The Q2 cohort scaled while getting cheaper
Clients who joined us in Q2 tell the scaling story more cleanly. QoQ, we scaled their combined spend 94%, from $239.2k to $464.1k, while CPA fell from $14.55 to $9.86, a 32% decrease.
Nearly doubling spend usually buys more expensive conversions, because each added dollar reaches less efficient inventory. Cutting CPA by a third at that scale is the payoff of building the account correctly before pushing budget into it.
Going Into Q4, Test Tighter Targets
Since Google's August 17 bidding update, we track actual CPA and ROAS against tCPA and tROAS more closely than ever. Our read is that the algorithm tends to stop once it hits the target you set, which leaves conversion opportunity on the table.
Across our book, we have seen multiple cases where asking for higher efficiency on a campaign already hitting its goal produced more revenue and more conversions at better efficiency. That runs against the usual assumption that a tighter target always costs volume.
How we run the test:
Pull every campaign where actual CPA beats tCPA, or actual ROAS beats tROAS, over a stable recent window.
Tighten the target in small steps, one campaign at a time.
Watch conversion volume and revenue alongside efficiency, since a tighter target that shrinks volume is a loss.
Roll the change out to similar campaigns only after the first test holds.
Do not run this test during BFCM week. Make the changes early in Q4 so each bid strategy has time to settle before peak traffic arrives.
Q3 made a strong case for fundamentals. Structure, keyword coverage, feed quality and testing drove the biggest wins in our book, while AI Max and Manual CPC both proved that a tactic's results in one account guarantee nothing in the next.
Before Q4 peaks, audit three things in your own account:
Check whether Manual CPC still earns its place on branded search, confirm AI Max is finding new queries instead of re-buying existing ones, and test tighter targets on campaigns already beating their goals.