Artificial Intelligence in Digital Advertising: What’s Real and What’s Noise

Artificial intelligence in digital advertising is no longer experimental. It is embedded in the platforms you already use every day, shaping how budgets are allocated, how audiences are targeted, and how creative is tested and served. The question for most marketing teams is not whether AI is involved in their campaigns. It almost certainly is. The question is whether they understand it well enough to use it deliberately rather than just letting it run.

That distinction matters more than most vendors would like you to believe.

I spent years managing large paid media accounts across multiple industries, including a period at iProspect where we grew the team from around 20 people to over 100 and moved from a loss-making position to one of the top five agencies in the market. A big part of that growth came from understanding how to operate at scale without losing commercial discipline. The platforms were simpler then, but the underlying challenge was identical: how do you make sure automation serves your strategy rather than replacing it?

What AI Is Actually Doing Inside the Ad Platforms

When marketers talk about AI in digital advertising, they often mean something vague and futuristic. The reality is more mundane and more immediate. Google’s Smart Bidding, Meta’s Advantage+ campaigns, and Amazon’s dynamic bidding tools are all machine learning systems operating in real time, adjusting bids based on signals that no human analyst could process fast enough to act on.

These systems are looking at signals like device type, time of day, browser history, location, and dozens of other contextual variables simultaneously. They are making micro-decisions on every auction. That is not a future capability. That is what happens every time someone clicks on a paid search result today.

Performance Max, Google’s campaign type that consolidates inventory across Search, Display, YouTube, Gmail, and Maps, takes this further. You provide the creative assets and the conversion objective, and the system decides where to show the ad, to whom, and at what bid. The machine is making decisions that used to sit with a media planner. It is faster. In many cases it performs better. But it is also considerably less transparent.

I remember the early days of paid search at lastminute.com, running campaigns where every keyword, every bid, and every match type was a manual decision. We launched a paid search campaign for a music festival and generated six figures of revenue within roughly a day from a relatively simple setup. The feedback loop was almost instant, and the control felt total. That world is gone. The platforms have taken the granular controls away, and in most cases the performance outcomes have improved. But understanding what is happening inside the black box has become harder, not easier.

If you want to go deeper on how AI is reshaping the broader marketing landscape, the AI Marketing hub at The Marketing Juice covers the full picture across channels, tools, and strategy.

Where Automation Genuinely Outperforms Human Optimisation

There are areas where AI-driven advertising optimisation is objectively better than what a human team can do. Acknowledging that is not defeatist. It is commercially sensible.

Bid management at scale is the clearest example. When you are running thousands of keywords across multiple markets with varying conversion rates, device splits, and dayparting requirements, a machine learning system will outperform a human analyst making manual adjustments. It has more data, it processes it faster, and it is not distracted by the other twelve things on its desk. The performance gains from switching to automated bidding are well documented by the platforms themselves, though you should always read those case studies with a degree of scepticism given who published them.

Audience targeting has also moved to a place where the platforms know more about behavioural intent than most advertisers can infer from their own first-party data. Meta’s lookalike audiences and Google’s in-market segments are imperfect, but they are built on data sets that dwarf what most brands have access to. The AI is doing something genuinely useful here, even if the mechanics are opaque.

Dynamic creative optimisation is another area where AI earns its place. Testing dozens of creative combinations manually across audience segments is slow and resource-intensive. Letting the system identify which headline, image, and call to action combination performs best for a given audience is a reasonable use of machine learning, provided you are feeding it creative that is worth testing in the first place.

The Semrush breakdown of how AI tools are being applied across marketing functions is a useful reference point for understanding where the technology is genuinely adding value versus where it is being oversold.

The Problem With Handing Over Control

Here is where I want to push back on the direction the industry is heading. The platforms have a strong commercial incentive to automate as much of the decision-making as possible. More automation means more data flowing through their systems, more inventory being accessed, and, in many cases, higher spend. That is not a conspiracy. It is just how their business model works.

The risk for advertisers is that automation optimises toward the objective you set, not necessarily the outcome you want. If you set a target cost per acquisition and the system hits it, you might assume the campaign is working. But if the customers being acquired have low lifetime value, churn quickly, or were already in your funnel through another channel, the CPA metric tells you almost nothing about commercial performance.

I have seen this play out in agency settings more times than I can count. A client is delighted because their CPA has dropped 30% since switching to automated bidding. Six months later they are puzzled about why revenue has not moved in proportion. The machine was optimising efficiently toward a metric that was not properly connected to business outcomes. The brief was wrong, not the algorithm.

The other issue is attribution. AI-powered campaigns spread spend across multiple touchpoints and inventory types, which makes it harder to understand what is actually driving conversions. Performance Max is a good example. It reports aggregate performance across all inventory, but it does not easily tell you whether your budget is working harder on YouTube or Search. That opacity is a problem if you are trying to make intelligent decisions about where to invest.

Moz has done useful work on how AI-generated and AI-influenced content is being evaluated for quality signals, which connects to the broader question of how platforms assess the credibility of what you are advertising. Their analysis of how LLMs interact with content signals is worth reading if you are thinking about how AI affects discoverability as well as paid reach.

Creative Is Now the Primary Competitive Variable

If the platforms are handling bid management and audience targeting with increasing sophistication, the area where human judgement still makes a decisive difference is creative. This is not a consolation prize. It is a genuine shift in where competitive advantage sits in digital advertising.

When I started in marketing, I taught myself to code because the agency I was at could not afford to build a new website. I needed a result and I found a way to get it. That instinct, finding where you can create an edge with the resources you have, is exactly the right frame for thinking about creative in an AI-driven advertising environment.

The platforms are commoditising targeting and bidding. Everyone using Performance Max is competing on the same inventory with systems that are broadly similar in capability. The creative you feed into those systems is one of the few remaining variables that is genuinely differentiated. A strong creative concept, a specific value proposition, a message that connects with a real human need, these things still matter enormously. The machine will distribute them more efficiently than you could manually. But it cannot generate them for you, at least not yet at the level of quality that moves markets.

AI-generated creative tools are improving quickly. HubSpot has covered the development of generative AI video tools in useful detail, and the capability gap between AI-assisted and human-led creative is closing faster than most creative directors would like to admit. But the brief, the strategic thinking that defines what the creative needs to do and why, remains a human responsibility.

The Ahrefs webinar on AI tools and their practical applications touches on this tension between automation and strategic input, which is relevant beyond SEO and applies directly to how paid media teams should be thinking about their role.

Signal Quality: The Input Problem Nobody Talks About Enough

Machine learning systems are only as good as the signals they receive. This is well understood in theory and routinely ignored in practice.

The signals that feed AI-powered advertising systems include your conversion data, your audience lists, your creative assets, and your campaign structure. If any of these are poor quality, the system will optimise efficiently toward a bad outcome. Garbage in, garbage out is not a new concept, but it applies with particular force to AI-driven advertising because the system has no way to know that the signals are misleading. It just follows them.

Conversion tracking is the most common failure point. If your tracking is capturing micro-conversions like page views or video plays as primary conversion events, the bidding system will optimise toward those events. You will get lots of them. They will not translate into revenue. The fix is not a technology problem. It is a strategic and analytical problem that requires a human to define what a meaningful conversion actually looks like for the business.

First-party data quality is the other area where most advertisers are leaving performance on the table. The platforms are increasingly reliant on advertiser-provided signals as third-party cookies diminish in value. If your CRM data is incomplete, your customer lists are stale, or your match rates are low, the AI has less to work with. Investing in first-party data infrastructure is not glamorous, but it is one of the highest-leverage things an advertiser can do right now.

Monitoring how your brand and your campaigns are being interpreted by AI systems is a growing area of practice. The Semrush overview of LLM monitoring tools is a useful starting point for understanding how to track AI-driven signals that affect your visibility and performance.

How to Stay in Strategic Control of AI-Powered Campaigns

The practical question for most marketing teams is not whether to use AI in their digital advertising. The platforms have already made that decision for you. The question is how to stay in strategic control of campaigns that are increasingly automated at the execution level.

Start with the objective. Before you configure any campaign, be precise about what you are trying to achieve and make sure that objective is measurable in a way that connects to business outcomes. A CPA target is only useful if you know what a customer acquired at that cost is worth to the business. If you do not have that number, get it before you set the target.

Invest in your creative and your brief. The brief is the document that tells the creative what to achieve and why. It is also, indirectly, the document that tells the AI what kind of performance to optimise toward. A weak brief produces weak creative, which produces weak signals, which produces weak results regardless of how sophisticated the bidding algorithm is.

Build a measurement framework before you launch. Decide in advance what you will look at to determine whether the campaign is working, and make sure those metrics are connected to real business value. Do not let the platform’s reporting dashboard define your success criteria. It will show you the metrics that make the platform look good.

Run structured experiments. AI-powered campaigns need a learning period, and they can be sensitive to changes during that period. But that does not mean you should never test. Build a cadence of controlled experiments, isolating variables where possible, so that you are generating your own evidence about what is driving performance rather than relying entirely on the platform’s attribution.

Keep humans close to the strategy layer. Automation handles execution well. It handles strategy poorly, because strategy requires context, commercial judgement, and an understanding of the business that no platform has access to. The marketers who will get the most from AI-powered advertising are the ones who are clear about which decisions belong to the machine and which belong to them.

Moz’s ongoing work on AI content and E-E-A-T signals is relevant here too, because the quality signals that affect organic search performance are increasingly connected to the credibility signals that affect paid performance. The two channels are less separate than they used to be.

For a broader view of how AI is reshaping marketing strategy beyond paid advertising, the AI Marketing section of The Marketing Juice brings together analysis across channels, tools, and the strategic questions that matter most right now.

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