Why AI Targeting Is Optimising Your Ads for the Wrong 5% of the Market
Nike spent $4.3 billion on marketing in 2024. Most of it went into performance marketing and programmatic display. The results: revenue fell 10% year on year, digital sales dropped 20%, and the company lost $25 billion in market value in a single day. Over nine months, the total damage reached $70 billion.
Nike had the most sophisticated AI-powered ad targeting on earth. It still shrank.
This isn't a Nike problem. It's a structural problem with how most businesses use AI in their marketing. And the research explains exactly why.
The precision trap every business falls into
The Ehrenberg-Bass Institute's Professor John Dawes established a heuristic that should reshape how every SME thinks about advertising: at any given moment, roughly 5% of your potential market is actively looking to buy. The other 95% will need what you sell eventually. They're just not shopping yet.
AI-powered ad platforms are extraordinarily good at finding that 5%. Google's Smart Bidding, Meta's Advantage+ audiences, Performance Max campaigns. All of it is engineered to identify purchase intent signals and serve ads to people showing them.
The problem isn't that this works. It does. The problem is that every competitor in your category is using the same tools, chasing the same 5%, at the same time.
The data shows what happens next. Average CPMs across Meta rose 20% year on year to $13.48 in 2026. In healthcare, CPMs jumped 70%. Meanwhile, click-through rates stayed flat even as costs rose. You're paying more to reach the same shrinking pool of in-market buyers, and so is everyone else.
Mark Ritson called this out directly: Nike's pivot to performance marketing was the textbook example of what happens when you mistake efficient targeting for effective marketing. The targeting got better. The brand got weaker. Preference for Nike in the US dropped 6% in six months, roughly 8 million fewer people who would choose Nike over competitors. Hoka and On Running walked into the gap.
Why the brand that comes to mind first wins the sale
Byron Sharp's research in How Brands Grow explains the mechanism. He calls it mental availability: the likelihood that your brand comes to mind when someone enters a buying situation. Not just awareness. Not whether they've heard of you. Whether you're the name that surfaces automatically when they think "I need a scaffolder" or "we should look at our marketing."
Mental availability is built through repeated, broad exposure over time. It's the accumulated effect of seeing a brand across different contexts, different moments, different messages. None of those individual exposures feel like they're doing much. But collectively, they determine which brands make the shortlist before anyone opens Google.
This is why AI targeting creates a blind spot. The entire architecture of performance marketing is designed to find people at the moment of decision. It does almost nothing for the months of passive exposure that determine which brands get considered at all. You're arriving at the auction after the buyer has already decided who they trust.
Ritson frames this as a false dichotomy. "Brand" and "performance" aren't two separate activities. They're two time horizons of the same job. Performance captures existing demand. Brand creates future demand. Cut the brand investment, and you're drawing down a bank account without making deposits.
Les Binet and Peter Field quantified this through the IPA effectiveness database: the optimal budget split for long-run growth is roughly 60% brand building, 40% activation. Most SMEs running AI-optimised campaigns sit at closer to 10/90 in the other direction.
The zero-click world makes this worse
Rand Fishkin's research at SparkToro adds another dimension. When he first measured zero-click searches in 2019, roughly 50% of Google searches ended without a click to any website. That number has climbed steadily since, accelerated by AI Overviews and featured snippets.
The implication for the 95/5 split: even when someone does enter the market and searches, they may never click through to your site. If the AI overview answers their question, if the map pack surfaces your competitor, if the knowledge panel resolves their need, your carefully targeted ad might never get seen.
Fishkin's argument is that brands need to build influence where buyers already spend time, not just where they search. YouTube, LinkedIn, local community groups, industry forums. The goal isn't always a click. It's being the name they already know when the buying moment arrives.
This converges with Sharp's mental availability concept from a completely different direction. Sharp arrived at it through consumer panel data. Fishkin arrived at it through search behaviour data. Both reach the same conclusion: if you're only visible at the point of purchase, you've already lost most of the battle.
For SMEs spending $2,000 to $5,000 a month on ads, this creates a strategic choice. You can spend it all fighting over the 5% in increasingly expensive auctions. Or you can allocate a portion to building the kind of visibility that makes the auction cheaper when you do show up, because the buyer already knows your name.
What the data actually says about broad vs narrow
Here's the counterintuitive finding. Even inside the AI-powered platforms, broader targeting is often outperforming narrow targeting on cost efficiency. Meta's Advantage+ Shopping Campaigns using broad targeting have reduced CPMs by up to 68% compared to interest-based targeting. The platforms themselves are telling you: stop narrowing. Let the algorithm find buyers you wouldn't have predicted.
This aligns with Sharp's penetration research. Growth comes from reaching more people in the category, including light buyers and non-buyers who will enter the market later. Trying to narrowly target only the most likely converters is the marketing equivalent of fishing in a barrel while the ocean sits next door.
| Approach | What it does | Cost trend | Growth effect |
|---|---|---|---|
| Narrow AI targeting | Finds the 5% in-market now | Rising (more competition for same pool) | Diminishing returns over time |
| Broad reach + AI optimisation | Reaches the 95% while AI optimises delivery | Lower CPMs, better efficiency | Builds mental availability for future demand |
| Brand + performance balance | Captures today's demand while creating tomorrow's | Stable or declining CPL over time | Compounding growth |
What to actually do about it
This isn't about abandoning performance marketing. Google Ads and Meta Ads are excellent at capturing demand that already exists. The in-market 5% are the most valuable people in your market right now. Keep reaching them.
But if that's all you're doing, you're building a business entirely dependent on people who are already searching. You're not growing the pool of buyers who will search next month. You're just getting better at competing for a fixed supply of demand while the cost of doing so rises every quarter.
Map your category entry points. What are the five to ten situations that cause someone to need what you sell? A scaffolding company's entry points might include starting a commercial build, a council requiring fall protection, or a current provider letting them down. Each one is a content opportunity that reaches the 95%. Use AI to create at scale, not just optimise. AI's real leverage for small business isn't smarter bidding. It's the ability to produce content across many more entry points than was feasible manually. Blog posts, videos, social content that covers the moments people enter your category. That's where the structural advantage lives. Let the platforms go broad. Stop over-constraining your audience targeting. Give the AI quality creative and clear conversion signals, then let it find buyers you wouldn't have predicted. The data says broad targeting with strong creative beats narrow targeting with average creative. Measure the right things. Mental availability building won't show up in your weekly Google Ads dashboard. Track branded search volume over time. Track direct traffic trends. Watch whether your CPL improves over months as more people arrive already knowing your name. These are the metrics that matter for long-run growth. Hold the 60/40 intent. You don't need to hit the ratio immediately. But have a plan for moving toward balance. Even shifting from 90/10 performance-heavy to 80/20 is a meaningful step. The businesses that build brand alongside performance will face lower acquisition costs, less platform dependency, and stronger pricing power over time.Nike is spending billions to reverse course. The lesson is available for free.
Further Reading
- How Brands Grow by Byron Sharp - The Ehrenberg-Bass Institute's evidence that brands grow through mental and physical availability
- The Long and the Short of It by Les Binet and Peter Field - The IPA's landmark effectiveness research on the 60/40 brand-to-activation balance
- Nike's $27.5 billion loss: A cautionary tale - Mi3's analysis of what went wrong when performance marketing displaced brand building
- Mark Ritson: Four big strategic mistakes Nike needs to reverse - Ritson's breakdown of the brand-vs-performance false dichotomy
- Rand Fishkin on brand influence in a zero-click world - Why visibility matters more than clicks in modern search
- AI ad automation raises costs without meaningful engagement gains - eMarketer data on the diminishing returns of AI-powered bidding
Dream Outcome is an Australian digital marketing agency helping SMEs grow through Google Ads, Facebook Ads, and Email Marketing.