Hosting my podcast, gives me a front-row seat to the unfiltered reality of our industry, the gritty, “in-the-trenches” reality shared by experts who manage millions in spend.
Last year, my guests, including Greg Finn, Christine “Shep” Zirnheld, Julie Friedman Bacchini, Andrew Lolk and Shawn Walker didn’t hold back. They dismantled “best practices,” called out platform biases, and highlighted exactly where the algorithms fail without human hands.
Here are the 10 most interesting (and sometimes uncomfortable) things my guests shared last year that you can take forward for 2026.
1. Google Is “Shaking The Couch Cushions” (And You’re The Couch)
We need to stop pretending Google’s incentives are perfectly aligned with ours. As Greg Finn and Christine Zirnheld from “Marketing O’Clock” pointed out, Google is, ultimately, a for-profit company, and while it remains an important advertising partner, its objectives don’t always perfectly align with what’s best for advertisers.
Finn put it perfectly: Have you ever noticed that the “Recommendations” tab always suggests raising your budget but never lowering it? That bias is literally built into the UI. With CPCs hitting record highs, “success” for the platform often just means “more revenue” for the shareholder.
And we’ve seen this play out in the data. Optmyzr analyzed more than 17,000 Google Ads accounts and found no consistent correlation between a high Optimization Score and strong performance. In fact, many of the best-performing accounts improved not by accepting Google’s recommendations, but by selectively rejecting them and focusing only on fixes that actually moved CPA, ROAS, or profit.
So, the takeaway is simple: Stop treating recommendations as gospel. Treat them as upsells, because the data shows that blindly following them doesn’t reliably help advertisers, but it does reliably help Google.
2. Automation Without Guardrails Is Just “High-Speed Waste”
The consensus from Shawn Walker from Symphonic Digital, Finn, and Julie Friedman Bacchini, President & Founder of Neptune Moon & Managing Director of PPC Chat, was unanimous: AI can execute, but it cannot strategize.
Walker noted that without strict conversion quality thresholds, Smart Bidding inevitably chases “cheap junk leads” because they are the easiest conversions to get. Meanwhile, Julie warned about “algorithm drift,” where a campaign slowly expands into irrelevant search terms because it thinks it’s being helpful.
Automation is necessary for modern account management, but that doesn’t mean “set it and forget it.” Your job isn’t moving bids anymore, it’s designing the right layers of automation and guardrails so the algorithms work, not the other way around.
I recently tested how well AI could diagnose a drop in conversions, and it confidently identified the campaign’s limited budget as the cause. The reasoning was simple: Some keywords were still receiving impressions while others weren’t, so the budget must be the bottleneck. But that didn’t make sense. Budget constraints usually affect campaigns broadly, reducing visibility across the board rather than selectively shutting off individual keywords.
When only certain keywords go dark, the more plausible explanation is a bidding issue. And if bidding is automated, that often indicates the algorithm deems those keywords as lower quality, resulting in lowered bids and ultimately, disappearing impressions.
The bigger point is this: AI often answers with confidence before it answers with accuracy. It can absolutely help you refine ad copy or strengthen relevance, but it still struggles to understand the nuanced and often counterintuitive interdependencies within a PPC account. In other words, it can assist with execution, but it’s not yet as reliable as a strategist.
3. The “Rule of 30” Is The New Law Of Gravity
One of the most practical takeaways of the year came from Walker. We often debate how much data Smart Bidding needs, but Shawn gave us the math:
You need ~30 conversions per campaign, per 30 days.
Not per account. Not across shared budgets. Per campaign. Below that threshold, the machine is just guessing. If you’re wondering why your small campaigns are volatile, it’s not bad luck; it’s bad math. You are starving the algorithm.
In Optmyzr’s 2024 study on the impact of bidding strategies on performance, we saw the same. 50+ conversions per month are ideal. 30+ is good, and anything less isn’t great. However, I would like to add one refinement to Shawn’s point. The real threshold isn’t “30 conversions per campaign,” but enough volume on the conversion goal/actions Smart Bidding is optimizing toward. Google’s systems can use broader, account-level conversion patterns to reduce data scarcity, and account-default goals and portfolio strategies are designed to expand the learning set beyond a single campaign.
What truly matters is having enough volume for the action you want Smart Bidding to optimize toward. If multiple campaigns are all working toward the same conversion event, they can effectively “pool” their learnings. In other words, campaigns don’t have to hit 30+ conversions individually as long as the underlying conversion action has enough aggregate volume for the system to learn and make reliable decisions.
4. “Soft Conversions” Are The Backbone Of SMB Success
So, what do you do if you can’t hit that magic number of 30? You have to feed the beast something else.
Guests heavily advocated for moving up the funnel. Walker detailed the necessity of “engaged visitor” signals, custom metrics like a user scrolling to a certain depth or spending time on site, fired only once per unique user to prevent inflation.
Whether it’s a PDF download, an add-to-cart, or a pricing page visit, these “soft” signals are no longer optional crutches; for smaller accounts, they are the only way to generate enough data density for Smart Bidding to function.
In other words, micro conversions still matter. They give Smart Bidding a richer sequence of intent signals to learn from: Did the user compare products? Did they view pricing? Did they return within 24 hours? Did they engage with interactive tools? In my experience, these micro-signals are what prevent smaller accounts from starving the algorithm and ultimately help it recognize high-quality users earlier in the journey.
5. SKAGs Are Finally, Truly Dead
If you are still using single keyword ad groups (SKAGs) in 2025, you are fighting a war that ended years ago. Bacchini was blunt: SKAGs have “run their course.”
The granular control we used to prize is now a liability. It fragments your data, making it harder for the AI to learn. Andrew Lolk, Founder at SavvyRevenue, backed this up, warning that over-segmenting campaigns destroys shared learnings. The winning structure for 2025 is radically simple: Consolidate until the data proves you need to separate.
What does that mean? Well, you should split campaigns when there are business reasons, like different bid targets, different promotions, etc. Put simply, you separate campaigns only when there’s a strategic reason, such as assigning different ROAS targets to products with different margins, or isolating seasonal inventory, like ski jackets, from evergreen categories like swimwear, so each can be optimized on its own performance curve.
And while single-keyword ad groups are outdated, single-theme ad groups (STAGs) have become the modern, more effective alternative. Instead of isolating each keyword, STAGs cluster queries that share the same intent and require the same message, giving Google enough data to learn without sacrificing relevance.
A better way to think about it:
A STAG isn’t just “all running shoes terms,” but it’s “all running shoes for distance training terms,” or “all waterproof trail-running shoes terms.” Each theme represents a specific user intent that warrants a specific ad and landing page combo
So, a more realistic STAG example might look like:
Theme: Long-distance running shoes
- “best long-distance running shoes”
- “marathon training running shoes”
- “long-distance running shoes men”
All different keywords, but they relate to the same core motivation, the same benefits to highlight, and the same landing page experience.
STAGs preserve the messaging control SKAGs once offered, but without the data fragmentation that hinders Smart Bidding from working at its best. They give you messaging precision while still feeding the algorithm enough volume to learn.
6. Stop Splitting Performance Max
Speaking of consolidation, Lolk had some strong words for how we manage Performance Max. A common mistake is splitting PMax campaigns by asset group, brand, or generic themes without a distinct ROAS target.
His take? “Splitting = Starving.”
PMax campaigns don’t share data well. If you split them, you force each new campaign to learn from scratch, requiring double the volume to stabilize. Unless you have a radically different ROAS target for a specific category, keep it together. And for the love of PPC, stop running “feed-only” PMax, he says. Just use Standard Shopping if you need that control.
7. Search Is Making A Quiet Comeback
In a surprising twist, we repeatedly heard that ecommerce brands have overemphasized PMax and Shopping, leaving money on the table in Search.
Lolk argued that Search is reclaiming its role as the high-intent workhorse because it offers what PMax cannot: diagnostic visibility and true messaging control. You can’t capitalize on a weather trend or a specific seasonal moment if you’re waiting for PMax to “learn” about it. Search lets you move fast, and it lets you control the landing page, a lever we’ve severely undervalued lately.
8. Your Competitive Advantage Is Now “Post-Click”
With Google automating bids, targeting, and even the creative process, what is left for us? Bacchini says the answer lies after the click.
Differentiation is the new battleground. If your offer is weak or your landing page is generic, no amount of bid tweaking will save you. Clients often dramatically underestimate their competitors and overestimate their own value propositions. As PPC pros, our value add is shifting from “technical setup” to “business consultancy,” fixing the offer, the positioning, and the user experience.
9. Generative AI Is Your New Junior Strategist
We moved past the “AI will write my ads” hype and got into real use cases.
- Zirnheld explained that AI has become her go-to tool for smoothing the communication gap between complex PPC work and client understanding. She uses it to draft clearer explanations, refine messaging, and spark creative concepts she can develop further. AI helps her accelerate the early stages, allowing her to spend more time on higher-value thinking.
- Walker described how AI has become a true technical force multiplier inside his workflow. He now uses it to write Google Ads scripts, build custom tools, generate and debug code, and automate tasks that previously required days of manual effort. AI effectively turns his ideas into working prototypes, allowing him to iterate faster and push the boundaries of what one PPC manager can build.
- Bacchini shared that AI has transformed how she researches competitors and analyzes positioning. Instead of manually combing through search results and landing pages, she can feed everything into AI and instantly see patterns, themes, and gaps. It gives her a strategic overview in seconds, helping her craft sharper messaging and understand where clients stand in a crowded landscape.
The consensus? AI won’t replace you, but an expert using AI will absolutely replace an expert who refuses to touch it.
In Silicon Valley, we used to lionize the idea of the 10x engineer, the kind of person who could out-code an entire team, see around corners in the architecture, and somehow ship things at a pace that felt almost unfair. But lately, the stories I’m hearing in my own network tell a different tale: Many of those “10x” engineers are starting to fall behind the so-called mediocre ones who are simply pushing the limits of what they can do with AI by their side.
And this no longer applies just to engineering. In every role, those who learn to partner with AI will outperform those who rely solely on talent and hustle.
10. The “Search” We Knew Is Disappearing
Finally, we touched on the existential shift. Shep mentioned she now uses Perplexity.ai for research more than Google. Greg Finn highlighted the instability of AI Overviews.
As I’ve been saying all year, we’re witnessing a dramatic shift from keywords to prompts. Search is no longer just about matching a query to an ad; it’s about connecting users who do complex prompts with solutions, and maybe showing an ad if that would be helpful.
In an AI-driven ecosystem, the “prompt” becomes the new keyword: a richer, more contextual signal that reflects not just what users type, but what they’re trying to accomplish. Advertisers who still think in terms of isolated keywords will fall behind; those who think in prompts, tasks, and intent paths will thrive.
The Bottom Line
The work of the modern PPC marketer continues to shift from pulling levers to thinking critically about the levers being pulled on our behalf. Automation is no longer optional, but neither is oversight. The winners this year were the advertisers who understood where algorithms shine, where they stumble, and where a human needs to step in with context that the machine simply doesn’t have.
And this evolution is far from over. As we head into 2026, I expect the debates on PPC Town Hall to get even more interesting. We’ll likely spend less time arguing about whether to adopt AI, and more time unpacking how to direct it, how to measure it, and how to prevent it from homogenizing every account it touches. We’ll explore what happens when prompts truly become the new keywords. And we’ll hear from practitioners who find creative, sometimes surprising ways to bend automation back toward profitability and strategy, rather than convenience.
If 2025 was the year we learned to tell the machine “No,” then 2026 may be the year we learn how to tell it “Do this and here’s why.” The marketers who thrive will be those who don’t just manage campaigns, but manage systems, using judgment, experimentation, and clear intent to guide increasingly powerful tools.
I’m looking forward to another year of unfiltered conversations on PPC Town Hall and to seeing what new hard truths (and opportunities) we uncover together.
More Resources:
- 20 AI Prompt Ideas & Example Templates For PPC (Easy + Advanced)
- 3 Ways AI Is Changing PPC Reporting (With Examples To Streamline Your Reporting)
- PPC Trends 2026
Featured Image: Anton Vierietin/Shutterstock
