RIP Your Google Ads Campaign Structure (2015–2026)

Okay, I need to talk about something that made me do an actual double-take this week. And I don't double-take easily — I've been in ecommerce long enough that most platform updates register somewhere between "mild inconvenience" and "ah, so we're doing this now." But Google just published guidance that essentially tells advertisers that the intricate, lovingly crafted campaign structures they've spent years perfecting are... actively hurting their performance.

That's right. The thing you were good at? Google would like you to stop doing it, please.

It's like spending a decade becoming the world's best horse-drawn carriage designer and then Henry Ford shows up and says "cool buggy, nerd." Except in this case, Henry Ford also built the road, owns the petrol station, and is charging you to drive on it.

The Architecture of Your Campaigns Is Apparently the Problem Now

For years, the mark of a skilled Google Ads manager was structural complexity. You segmented by match type, device, audience, product category — the whole beautiful, neurotic symphony. You had spreadsheets with colour-coded tabs. (Don't pretend you didn't.) That complexity was the moat. It was the reason clients paid you instead of just hitting "boost post" and praying.

Google is now saying: tear it all down.

Their argument goes something like this: the more you segment your campaigns, the less data each one gets. The less data each one gets, the worse the machine learning performs. So consolidate everything, feed the algorithm, and trust the system.

Which — and I cannot stress this enough — is Google asking you to trust Google. The company that once told us all to build exact match keyword empires and is now saying "actually, forget all that, just give us your budget and we'll handle it." My partner does this thing where they rearrange all the kitchen cupboards every six months and act like the new layout was always the obvious answer. Google Ads strategy now has that same energy.

Keywords Are Dead. Long Live... Vibes?

Here's where it gets properly weird. Google's search auction doesn't really work the way most of us were trained to think about it anymore. The old model was beautifully simple: user types query, system matches keywords, auction fires, you pay. Clean. Logical. Satisfying in the way that colour-coded spreadsheets are satisfying.

The new model is what Google calls "intent inference," which sounds like something a philosophy student would put on their CV. Instead of matching your keyword to a query, the system tries to figure out what the person actually wants and serves ads against that inferred need.

So someone searching "why is my pool water green?" isn't looking for pool chemicals. They're looking for answers. But Google's AI goes "ah, but they will need pool chemicals" and serves those ads anyway. It's like going to the doctor for a headache and they hand you a leaflet for a mattress store. Technically relevant. Spiritually unnerving.

For those of us selling products online, this rewires campaign planning from the ground up. We've been building strategies around what people search for. Now we need to build strategies around what people are trying to solve. Which is a much squishier, more existential question that I definitely didn't need at 11 PM on a Tuesday.

The Data Hunger Games

Google's push toward consolidation isn't just philosophical navel-gazing. There's a hard data economics argument underneath, and annoyingly, it kind of makes sense.

Machine learning needs volume. A campaign getting 15 conversions a month can't generate meaningful signals — it's basically trying to learn Italian from a phrasebook with half the pages missing. When you fragment your budget across dozens of tightly segmented campaigns, each with its own bidding logic and audience restrictions, you're essentially putting the algorithm on a starvation diet and then complaining it looks thin.

Google's argument is that one broad campaign getting 150 conversions will outperform ten campaigns getting 15 each, even if those ten campaigns feel more controlled. The AI can spot patterns across user behaviour, contextual signals, and real-time auction dynamics that no human segmentation strategy could replicate.

And here's the uncomfortable bit: for a lot of advertisers, they're probably right.

But — and this is a big "but" that keeps me up at night alongside all my other "buts" (my brain is basically a warehouse of anxious caveats at this point) — Google's own guidance suggests you need roughly 30 conversions in 30 days before automated campaigns can even start optimising effectively. If you're a smaller ecommerce brand doing maybe 20 conversions a month, you're caught in what practitioners are calling a "scissors gap": you don't have enough volume to train the algorithm, but you can't build volume without effective campaigns.

It's the advertising equivalent of "you need experience to get a job, but you need a job to get experience." Brutal. Familiar. Structurally unfixable by just consolidating harder.

The Trust Problem Nobody Wants to Name

Let's talk about the elephant doing a tap dance in the middle of the room.

Google's optimisation objectives are not the same as your optimisation objectives. The algorithm maximises for whatever metric you tell it to — conversions, conversion value, whatever. But how it gets there, and where your budget goes in the process, is substantially within Google's discretion under these automated campaign types.

You're essentially handing your credit card to someone and saying "buy me something nice" and then they come back with receipts you can't fully read. Maybe they did great! Maybe they spent 40% of your budget on placements you'd never have approved if you'd seen them. With Performance Max in particular, the reporting is... let's charitably call it "curated."

This isn't conspiracy thinking. It's just incentive alignment — or the lack thereof. Google makes money when you spend money. Google's AI decides how your money gets spent. Google also grades its own homework. If your internal alarm bells aren't at least gently tinkling at this arrangement, I'd like some of whatever calming tea you're drinking.

So What Do We Actually Do?

Look, I'm not saying fight the tide. The direction of travel is clear, and trying to maintain 2019-era campaign architecture in 2026 is going to get increasingly expensive and increasingly pointless. Google's feature development is all pointing one way, and that way is "let the machines drive."

But our role changes. If the algorithm handles tactical optimisation, the human contribution shifts upstream — to audience strategy, creative direction, first-party data quality, and (this is the big one) the quality of product data you're feeding into the system.

That last point matters enormously, especially if you're selling on both Google and Amazon. The signals these AI campaigns depend on — rich product attributes, accurate categorisation, quality images, detailed descriptions — are the same signals feeding Amazon's algorithm and every other AI-mediated discovery platform. Your product data isn't just catalogue hygiene anymore. It's the foundational input for every automated system deciding whether your product gets shown or buried.

On the keyword side, the practical guidance from people who've actually tested this: keep exact and phrase match for brand defence and high-intent bottom-funnel terms. Let broader match types and Performance Max handle the exploratory, upper-funnel stuff. Think of segmentation not by match type but by intent stage — where someone is in their decision process, not the specific words they typed.

The Bigger, Slightly Terrifying Pattern

Here's the thing that really gets me: this isn't just a Google story. It's the same story playing out everywhere simultaneously.

Meta did it with Advantage+ campaigns. Amazon's doing it with automated targeting and dynamic bidding. Every major ad platform is systematically reducing advertiser control in favour of algorithmic automation, justified by performance claims that are conveniently difficult to independently verify.

The common thread? Each platform's business interests are served by advertisers feeding more data into more consolidated systems. Less friction for them. More "optimisation latitude" (a phrase that should come with a warning label). And the expertise that made people valuable in digital advertising for the last decade — building complex, precisely controlled campaign architectures — is being actively deprecated by the platforms that once rewarded it.

The new expertise looks different: understanding what signals to feed automated systems, structuring measurement to account for AI-mediated attribution, and retaining strategic visibility when tactical decisions are increasingly made by machines you can't fully interrogate.

• $285B vanished in a day as Anthropic’s Claude agents triggered a “SaaSpocalypse,” repricing UI-driven SaaS and signaling a shift from per-seat software to API-first, agent-run workflows rewriting enterprise economics.

• Claude Opus 4.6 now builds and ships full Chrome extensions in minutes via Shipper, from code to privacy policy, turning browser dev into a $0.11 prompt and putting no-code automation on steroids.

• Meta AI teases Avocado models, Manus browser agents, MCP connectors and scheduled tasks, a full-stack push to close the gap with OpenAI and Anthropic and evolve into a true multi-agent platform.

• Gemini is testing ChatGPT chat imports, letting users upload old threads and keep context, but with a catch: imported history becomes part of Google’s model training ecosystem.

✒️ Fora - Artificial intelligence designed for the C-suite.

🤖 Shortwave AI Assistant on iOS - Write, search, and analyze your emails with AI on the go.

🧑 Taskade Custom AI Agents - Specialized tools designed to automate routine activities like research, task management, and content creation

📝 Superlist - Create to-do lists, capture thoughts or detailed notes, assign tasks to teammates, and more

Copywriter: Helps write engaging ad copy to drive action.

🖥️ Stori AI: Create and publish branded content effortlessly. Reach your audience across multiple channels and make a lasting impression.

😍It was great, all I needed

🙂Good, not great

🙄It sucked