Creative Is the New Targeting (And Most Advertisers Still Don't Believe It)

Manual audience targeting is not dead. It has moved. It now lives inside your creative.

That is the most accurate way to describe what has happened to paid media over the past few years. Meta has shifted toward broad and default audience settings, with its delivery system using engagement signals from the ad itself to decide who sees it next. Google is auto-upgrading accounts to AI Max in September 2026. The platforms are making the decision for you, and the main input you still control is your creative.

Meta put it plainly: the focus has shifted "from niche targeting to creative diversification as the best lever to find the most relevant audiences." That is not a practitioner's opinion. That is the platform telling you how its system works.

For many advertisers, this implication is uncomfortable. Years of targeting expertise, audience segmentation, and persona-driven campaign structures may now be worth less than a strong hook in the first three seconds of a video.

What the Algorithm Is Actually Doing

Chapter 28 of Never Always, Never Never draws a useful distinction between structured and unstructured learning. It is the key to understanding why this shift happened.

Google's Smart Bidding is a structured learning system. It trains on labeled data: conversions, bids, placements, keywords. Feed it good conversion signals and it learns to predict which inputs produce the best outputs. The system is powerful, but bounded. It optimizes around what it has already seen.

Meta's audience discovery tools use unstructured learning. The algorithm explores user data without predefined labels, finding patterns that no human would have thought to look for. When Facebook's algorithm was given freedom to explore signals outside typical buyer personas, it successfully connected high-value customers in ways traditional segmentation never would. The personas were not just unhelpful. They were in the way.

Most advertisers miss this part. The system is not looking at your audience settings and then finding your ad. It is looking at your ad and then finding your audience. The creative is the signal. The hook angle, the copy framing, the offer structure, the emotional tone: these are not just persuasion tools. They are targeting inputs that tell the algorithm what kind of person is likely to engage.

When your creative library is thin, the algorithm has nothing to work with. It cannot differentiate. It cannot learn. It lands in what the book describes as a local optimum, a decent result based on the path it happened to take, not the best result available.

The Liquidity Problem Nobody Is Talking About

Drawing from a 2019 Interactive Advertising Bureau paper, the book introduces a concept worth understanding: liquidity. In a liquid campaign environment, every dollar flows to the most valuable impression. The machine reads the terrain. The human steps back.

Liquidity has four dimensions: placement, audience, budget, and creative. The fourth one is where most advertisers are leaving performance on the table.

Creative liquidity is achieved when the system can test and choose between meaningfully different assets. Not different colorways of the same static image. Different concepts. Different emotional registers. Different formats. Short punchy video alongside a longer explanatory one. A product demo next to a customer story. A System 1 ad designed to land in a distracted scroll alongside a System 2 ad built for someone actively evaluating options.

Restrict any of the four dimensions and you reduce what the algorithm can learn. Restrict creative liquidity and you are essentially running a single audience hypothesis over and over, hoping it is the right one.

Practitioners who have tested broader creative variety against narrow creative libraries consistently find that audience discovery improves when the algorithm has more signals to work with. Platform guidance confirms the direction. Advantage+ setups are now the default for many campaign types on Meta. The audience input is advisory. The creative input is determinative.

The Attention Spectrum Explains Why Volume Is Not Enough

Here is where the "just make more ads" advice falls apart.

Volume matters. But volume of what? A hundred variations of the same concept does not give the algorithm a hundred different signals. It gives it one signal, repeated noisily.

Chapter 12 of the book makes the case that consumers operate across an attention spectrum. Most of the time, people are in a System 1 state: automatic, emotional, low-effort processing. They are scrolling half-asleep, commuting, waiting in line. Occasionally, they shift into System 2: deliberate, rational, evaluative. A person actively shopping for insurance is in a very different mental state than someone who just happened to see a Geico ad while watching TV.

Daniel Kahneman's framework in Thinking, Fast and Slow explains why a single creative approach cannot serve both states. A detailed product comparison works for System 2. It is invisible to System 1. A quick, emotionally resonant image works for System 1. It does nothing for someone who is ready to make a considered decision.

Algorithms know which state a user is likely in. They track scroll speed, time on platform, recent engagement patterns, whether sound is on or off. What they cannot do is match a user's current state to a creative that fits it, if you only gave them one type of creative.

Meaningful variety matters for this reason, not cosmetic variety. Different tones. Different structures. Different messages for different moments. Building a creative system that can work across the full attention spectrum is the goal, not just one end of it.

Conventional wisdom that shorter is always better also deserves scrutiny. People have not lost the capacity for focus. They have become more selective about where they invest it. Someone will skip your ad in half a second and then watch a three-hour podcast without blinking. The question is not length. It is whether the creative earns the attention it asks for.

Personas Are Not the Input. Behavior Is.

The pattern of algorithms outperforming human audience assumptions is worth sitting with. When a platform's algorithm is constrained by human assumptions about who the customer should be, performance can suffer. When those constraints are removed, the algorithm finds its own clusters, cohorts that have little obvious in common but behave similarly in ways that matter.

At AdVenture Media, this pattern has appeared repeatedly across client categories. The audience that converts does not always match the audience you would have targeted manually.

This connects directly to what the Ehrenberg-Bass Institute has documented about buyer personas: detailed customer profiles often mislead because they represent the heavy buyer, not the broad market. Byron Sharp's research shows that brand growth comes primarily from reaching light and non-buyers, the people who do not fit your persona but might buy once. Manual targeting actively excludes them. Algorithm-driven delivery with broad settings finds them.

Your ads still need to speak to someone. But that someone is less specific than persona-driven strategy assumes. A hook that resonates with your target customer often resonates with a much wider group, and the algorithm will find that group if you let it.

The Contrarian Case for Targeting (That Actually Agrees With the Thesis)

Here is the honest version of this argument: targeting has not disappeared. It has been absorbed.

Platforms still use audience signals. They still use behavioral data, purchase history, and contextual information. What has changed is who controls the input. Manual targeting layers told the system who to reach. Now the system decides who to reach, and it uses your creative as one of the primary signals for that decision.

Conversion signal quality still matters enormously. A tracking pixel placed on the wrong page, a flawed attribution setup, a campaign structure that starves the algorithm of data: these will all undermine creative quality. The book describes the "dangerous AI" quadrant: high confidence, low accuracy, where the system is confidently optimizing for the wrong outcome because the data you fed it was wrong. No amount of creative variety fixes a broken measurement setup.

Campaign structure matters too. Budget liquidity requires consolidation. Fragmenting spend across many small campaigns with individual budgets starves each one of the signal volume needed to exit the learning phase and make confident predictions. This is not a creative problem. It is a structural one.

Creative is now the primary lever available to you, because the others have been largely handed to the machine. Use that lever well.

What This Means in Practice

Several things follow from this, and they conflict with how most teams currently operate.

Stop treating creative testing as a finishing touch. If the creative is now doing the targeting work, creative strategy belongs earlier in the planning process, not as an execution detail after the audience settings are locked.

Diversify concepts, not just executions. Swapping a headline or changing a background color does not give the algorithm a meaningfully different signal. Different concepts, different emotional angles, different formats: that is what creates genuine creative liquidity.

Evaluate performance at the campaign level. The book's breakdown effect analysis is directly relevant here. Meta allocates budget to placements in ways that optimize aggregate performance, not individual line-item appearance. A creative that looks expensive in isolation may be the right choice given what else has already been captured. Judge the portfolio, not the component.

Build a creative system, not a creative. Aim for a library that can serve the full attention spectrum across different contexts, mindsets, and moments. Some assets should be fast and frictionless. Others can earn deeper attention from people who are genuinely evaluating. Both have a role. The algorithm will route accordingly.

Keep feeding the algorithm accurate signals. Creative variety is only half the equation. Conversion data sent back to the platform determines whether the system learns the right lessons from what it serves. Garbage in, garbage out, regardless of how good the creative is.

Strategy First. Then Let the Machine Work.

Creative-as-targeting is real, and the direction is not reversing. Google is moving further toward automated campaign management. Meta's Advantage+ defaults are already standard. Both platforms are clear about where they are going.

But Never Always, Never Never argues throughout, and Chapter 28 states explicitly: platform optimization is always secondary to strategy. Machines amplify whatever you give them. Give them weak creative, a muddled value proposition, or a product with no real differentiation, and the algorithm will spend your budget proving that out with extraordinary efficiency.

Creative is the new targeting. That does not make it the new strategy.

Brand vs performance budget allocation still matters. Mental availability still needs to be built. Les Binet and Peter Field's 60/40 principle, documented in the IPA DataBank, holds that brand investment and performance investment need to work together. That does not change because Meta automated audience delivery. If anything, the case for brand investment gets stronger when your targeting is increasingly out of your hands and your main competitive lever is the quality and resonance of what you put in front of people.

Winning advertisers are not the ones who make the most creative variants. They are the ones who combine creative volume with creative intelligence: understanding the attention spectrum, building assets that can work across it, and feeding the algorithm the clean conversion signals it needs to learn accurately.

Get the strategy right. Then give the machines something worth amplifying.

---

Patrick Gilbert covers how ad platform AI learns, including the breakdown effect and the four quadrants of confidence versus accuracy, in Never Always, Never Never: Strategic Marketing in an AI World.