AI in Advertising Is Changing Fast. Here Is What Matters

AI in advertising is no longer a horizon story. It is operational, embedded in bidding systems, creative workflows, audience targeting, and measurement frameworks across the industry right now. The question is not whether AI will reshape advertising, it already has. The question is which changes will compound into genuine competitive advantage and which are noise dressed up as progress.

What follows is not a vendor pitch or a breathless forecast. It is a grounded look at where AI is creating real commercial value in advertising, where the hype outpaces the evidence, and what senior marketers should actually be paying attention to.

Where AI in Advertising Is Already Doing Real Work

Paid search was the first place I saw machine learning genuinely outperform human management at scale. When I was running performance marketing at iProspect, we were managing campaigns across hundreds of clients and hundreds of millions in spend. The manual optimisation work, adjusting bids by device, time of day, audience segment, match type, was enormous. Smart Bidding, even in its early iterations, did not just reduce that workload. It made decisions faster and with more signal than any human team could process in real time.

That is not a small thing. The compounding effect of better bid decisions across thousands of auctions a day is significant commercial value. And it is now table stakes. Every serious advertiser is running some form of automated bidding. The differentiation has shifted upstream, to the quality of the conversion signals you feed the algorithm, the campaign structure you build around it, and the creative you pair with it.

Audience targeting is the second area where AI has moved from experiment to infrastructure. Lookalike modelling, predictive audiences, and real-time behavioural signals are now core to how most programmatic and social campaigns find their highest-value customers. The platforms have invested heavily here because better targeting means better outcomes for advertisers, which means more spend on the platform. The incentives align, which is why the technology has matured quickly.

If you want a clear-eyed overview of how AI tools are being applied across the marketing stack right now, the AI Marketing hub on The Marketing Juice covers the landscape in detail, from workflow integration to tool selection to where the genuine limitations are.

What Is the Real Impact of AI on Creative in Advertising?

Creative is where the conversation gets more complicated, and more interesting.

There is a version of the AI creative story that is mostly about cost reduction. Generate more assets faster, reduce reliance on expensive production, scale personalisation across markets. That is real and it matters, particularly for performance creative where volume of testing is directly correlated with finding what works.

But the more important shift is what AI does to the testing cycle. When I started in advertising, running a meaningful creative test required budget, time, and production resource. You might test four or five variants in a campaign if you were disciplined about it. Now the constraint is not production, it is the quality of your creative hypothesis. AI can generate the variants. The scarce resource is knowing what to test and why.

That is a genuine strategic shift. Teams that understand their audience well enough to generate sharp creative briefs will extract more value from AI creative tools than teams that use them to produce volume without direction. Semrush’s breakdown of AI copywriting approaches is worth reading if you want to understand the current state of AI-assisted copy generation, including where it holds up and where it still needs a strong human brief to produce anything useful.

The platforms are pushing hard into AI creative too. Google’s Performance Max, Meta’s Advantage+ creative, and similar products are essentially asking advertisers to provide raw assets and letting the machine assemble and optimise the combinations. The results are genuinely mixed. When the underlying assets are strong and the conversion signals are clean, these products can perform well. When the assets are generic or the signals are weak, the machine optimises toward the wrong outcomes efficiently. That is a worse outcome than a poorly managed manual campaign, because at least with the manual campaign you can see what is happening.

Is AI Going to Fix Advertising Measurement?

This is the area I find most genuinely promising, and also the one where the gap between potential and current reality is widest.

Advertising measurement has been broken for a long time. Last-click attribution was always a fiction. Multi-touch attribution models are better but still fundamentally flawed because they assign credit based on correlation rather than causation. I spent years presenting attribution data to clients that I knew was a perspective on reality rather than reality itself. The clients who understood that were better at making decisions. The ones who treated the dashboard as ground truth made expensive mistakes.

AI-driven measurement approaches, particularly incrementality testing and media mix modelling, offer something more honest. Incrementality testing asks the right question: would this conversion have happened without this ad? Media mix modelling, when built on sufficient data and run with appropriate rigour, can give you a more defensible view of channel contribution than any click-based attribution model.

The challenge is that both approaches require data volume, statistical discipline, and a willingness to accept less precise but more accurate answers. Many marketing teams are not set up for that. They want a dashboard that tells them which channel drove which sale, and they want it to update in real time. That is an understandable desire and a completely unrealistic one. AI can improve the quality of measurement significantly, but it cannot make a fundamentally uncertain problem into a certain one.

What it can do is make better approximations faster, surface anomalies in performance data that human analysts would miss, and reduce the time between a campaign change and a reliable read on its impact. That is commercially valuable, even if it is less satisfying than the false precision of a last-click model.

How Will AI Change the Relationship Between Advertisers and Platforms?

This is a question the industry is not talking about enough.

The major advertising platforms, Google, Meta, Amazon, and the programmatic ecosystem around them, are all building AI layers that sit between the advertiser and the audience. They are asking advertisers to trust the machine with more of the campaign, to provide goals rather than tactics, outcomes rather than settings. In return, they promise better performance.

Sometimes that promise is kept. I have seen Performance Max campaigns outperform tightly managed manual campaigns in the right conditions. But the conditions matter enormously, and the platforms have an obvious commercial incentive to encourage broader adoption of their AI products regardless of whether those conditions are met in any specific case.

The risk for advertisers is a gradual loss of visibility and control. When the platform controls targeting, creative assembly, bidding, and placement, the advertiser’s leverage in that relationship diminishes. You become dependent on a black box that you cannot audit, cannot fully understand, and cannot easily exit. That is a strategic vulnerability, particularly for brands where advertising is a significant driver of revenue.

The answer is not to reject platform AI tools. They are too effective in the right contexts to ignore. The answer is to maintain enough first-party data, measurement capability, and internal expertise to evaluate what the platforms are telling you rather than simply accepting it. Ahrefs has covered the evolving AI tool landscape with useful rigour if you want to understand how the broader ecosystem is shifting, not just the platform side.

What Does AI Mean for the People Who Work in Advertising?

When I grew the team at iProspect from around 20 people to over 100, the roles we were hiring for changed significantly over that period. The junior analyst jobs that had been about pulling data and building reports were being automated or semi-automated. The roles that grew were the ones that required judgment, client relationship, strategic thinking, and the ability to translate data into decisions.

AI is accelerating that pattern. The tasks that are most vulnerable are the ones that are high volume, rule-based, and do not require contextual judgment. Writing first drafts of ad copy, building audience segments from defined parameters, generating performance reports, adjusting bids within defined guardrails. These are not disappearing overnight, but they are becoming less labour-intensive, which means teams need fewer people to do them, or the same number of people can take on more.

The tasks that are becoming more valuable are the ones AI cannot do well. Understanding why a client’s business is underperforming and what advertising can realistically contribute to fixing it. Building the trust required to have an honest conversation about measurement limitations with a CFO who wants certainty. Knowing when the data is telling you something real and when it is an artefact of how you set up the tracking.

That last one matters more than most people acknowledge. I have seen campaigns that looked like they were performing well on every dashboard metric and were quietly cannibalising organic revenue. I have seen campaigns that looked mediocre in the attribution model and were generating significant incremental revenue that the model was not capturing. The ability to hold that kind of scepticism about your own data is a human skill, and it is one that becomes more important as AI generates more data more quickly.

There is also a craft dimension to this. Good advertising requires genuine understanding of what makes people respond to a message. AI can analyse patterns in what has worked historically and generate variants based on those patterns. It cannot feel the difference between copy that is technically correct and copy that lands. That gap is real, even if it is narrowing.

Where Is the AI Advertising Hype Outpacing the Reality?

A few places are worth calling out directly.

Hyper-personalisation at scale is one. The promise is that AI will allow every consumer to see a version of your advertising that is perfectly tailored to their individual context, preferences, and moment. The reality is that the data required to do this well is difficult to collect, increasingly restricted by privacy regulation and platform policy, and often less predictive of behaviour than simpler demographic or behavioural signals. The gap between what is theoretically possible and what is practically achievable for most advertisers is large.

Autonomous campaign management is another. The idea that you can set a revenue target, give the machine a budget, and walk away is appealing. It is also, in most cases, a recipe for expensive surprises. The machine optimises toward the signals you give it. If those signals are imperfect, and they almost always are, the machine will find the most efficient path to the wrong outcome. Human oversight is not optional, it is the thing that keeps the system honest.

Predictive creative is a third area where the claims often exceed the evidence. Tools that promise to tell you which creative will perform best before you run it are solving a genuinely hard problem, and most of them are not solving it as well as they claim. Pattern matching on historical data can give you useful signals about what has worked in similar contexts. It cannot reliably predict how a specific piece of creative will perform in a specific auction environment against a specific competitor set. The Semrush analysis of AI-driven optimisation approaches is useful context here, particularly on the gap between AI-assisted and AI-autonomous decision making.

None of this means these tools are without value. It means you should evaluate them against what they actually do, not what the vendor deck says they do.

What Should Advertisers Actually Do About AI Right Now?

The early part of my career taught me something that has stayed with me. When I could not get budget for a new website at my first marketing job, I taught myself to code and built it. Not because I wanted to be a developer, but because the problem needed solving and waiting for someone else to solve it was not an option. AI is a similar moment for a lot of marketing teams. The tools are available. The question is whether you are going to learn how to use them or wait for someone else to figure it out first.

Practically, that means a few things.

First, audit where your team is spending time on tasks that AI can do faster or better. Creative iteration, audience segmentation, performance reporting, and keyword research are the obvious starting points. Moz’s work on AI-assisted content briefs is a good example of how AI can compress the research and planning phase without removing the strategic judgment that makes the output useful.

Second, invest in your first-party data infrastructure. Every AI tool in advertising, whether it is a platform’s bidding algorithm or your own predictive model, performs better with clean, rich, first-party signals. This is not a new insight, but it is one that becomes more urgent as third-party data becomes less reliable and more restricted. The teams that have invested in this over the past few years are now in a significantly stronger position than the ones that did not.

Third, build measurement capability that can evaluate AI-driven campaigns honestly. That means incrementality testing, not just attribution. It means being willing to run holdout groups and accept that some of your spend is not generating incremental return. It means having the internal credibility to present that finding to leadership without it becoming a political problem. Ahrefs has explored how AI is changing the measurement landscape with particular focus on search, which is where the signal quality issues are most acute right now.

Fourth, do not outsource your strategic thinking to the machine. AI can tell you what has worked, what patterns exist in your data, and what variants are worth testing. It cannot tell you whether you are solving the right problem, whether your positioning is differentiated enough to matter, or whether the brief you have given it is worth executing. Those are human jobs, and they are the ones that determine whether AI makes you more effective or just more efficient at doing the wrong things faster.

If you want to go deeper on how AI is reshaping the broader marketing function, not just advertising, the AI Marketing section of The Marketing Juice covers the full picture, from generative tools to workflow integration to what the technology genuinely cannot do yet.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions