AI Is Not the Expert: Why Human Judgment Still Matters in PPC
I’ve been working in paid media since 2008, and if there’s one thing I’ve learned, it’s that this industry never stays still.
Today, alongside my work as Paid Media Editor at Search Engine Land, I run PPC Live, its community and events, and the PPC Live podcast. That means I spend a lot of my time talking to people working at the sharp end of paid media: hearing what is working, what is changing and, perhaps more importantly, what is going wrong.
And right now, almost every conversation eventually leads to AI.
There is enormous excitement around what AI can do for PPC. It can analyse information, help manage campaigns, generate creative, assist with reporting and make previously time-consuming tasks significantly faster.
But I think we need to remember something important:
AI is nowhere near as mature as we sometimes pretend it is.
And having access to AI does not remove the need for expertise. If anything, it makes expertise more important.
AI still needs someone who knows what “good” looks like
Recently, I came across two posts that caught my attention. One came from an SEO strategist and another from PPC veteran Kirk Williams.
They were effectively making the same argument from two different sides of search: AI works when somebody who understands the fundamentals is steering it.
That overlap is important.
SEO and PPC professionals have historically operated in separate worlds, but our work is much more interconnected than we sometimes admit. Customers certainly don't think about search results in terms of internal marketing departments.
They search. They scroll. They click.
They don't necessarily care whether the result they chose came from PPC or SEO.
That is why I've always encouraged PPC and SEO teams to talk to each other. Go for coffee. Compare notes. Share what you're seeing.
Both disciplines conduct keyword research. Both think about competitors. Both care about copy, landing pages, search behaviour and intent. And increasingly, both are dealing with the same questions around automation and AI.
The same principle applies when adopting AI tools.
An experienced PPC professional can look at an AI-generated recommendation and recognise that something doesn't make sense. Someone without those fundamentals may simply assume the machine knows better.
That's where I see the risk.
There is no universal AI solution
We're entering an era that reminds me slightly of the old snake-oil salesman.
You see posts promising that if you connect this AI tool, use this prompt or build this workflow, suddenly everything becomes easier.
But marketing has never worked like that.
Paid search itself isn't right for every business. You can't simply tell every company to launch Google Ads and expect the same outcome.
Why would AI be any different?
Every business has different customers, margins, objectives, products, buying journeys and internal challenges. The person demonstrating an incredible AI workflow may have spent years developing the knowledge required to build it.
Then someone much less experienced copies the workflow and wonders why they aren't getting the same results.
The problem isn't necessarily AI.
The problem is believing the tool replaces the knowledge required to use it.
We need to understand AI's role in the customer journey
We've been through versions of this problem before.
Think about display advertising.
For a long time, marketers could look at display campaigns and think they weren't performing because their direct conversion metrics weren't comparable with search.
Then you switched display off.
Suddenly, search performance changed.
Eventually, we became better at understanding the role display played within the wider funnel. We stopped expecting every channel to behave identically and started asking what job each channel was actually doing.
We need to do the same thing with AI.
Rather than rushing to implement every new capability, we need to understand where AI sits in the journey, what it influences, which metrics matter and how it interacts with other channels.
That requires marketing knowledge.
It requires understanding attribution.
It requires understanding customers.
And it requires understanding the fundamentals before introducing another layer of technology.
Goal alignment is still one of our biggest problems
If I had to identify one area where things go wrong most often, it would be goal alignment.
AI cannot magically understand what matters most to a business.
It might see revenue increasing and conclude that performance is improving. But perhaps the company is about to run out of stock for the product driving that revenue. Perhaps another product has much better margins. Perhaps the leadership team is deliberately trying to grow a different part of the business.
The platform doesn't inherently know that.
The marketer needs to know it.
That's why reporting needs to become less about presenting numbers and more about having conversations.
Instead of simply telling a client that revenue, conversion rate or ROAS increased, I want to know:
What has changed in the business?
What matters to you right now?
Which products are strategically important?
What are your salespeople seeing?
What is happening to profitability?
What is the CFO concerned about?
That context changes how we interpret marketing performance.
AI can help us process information. It cannot replace the conversation that tells us which information matters.
Reporting should speak the language of the business
For years, digital marketers have loved our acronyms.
CTR. CPC. CPA. ROAS.
But your client may not care about most of them.
That doesn't mean you need to “dumb down” your reporting. It means you need to speak the client's language.
If profit is the issue, talk about profit.
If average order value matters, talk about average order value.
If there is one commercial metric that leadership really cares about, understand what it is and connect your marketing activity to it.
This also means agencies sometimes need access to information clients are reluctant to share.
I've heard examples of marketers working inside businesses who still struggled to access the financial information required to properly evaluate performance.
My advice is to start that conversation early.
Explain why you need the information. Explain how it will improve decision-making. Agree to the necessary privacy protections.
And don't necessarily expect a yes the first time.
Keep demonstrating why having that context would allow you to make better decisions.
## AI has not solved conversion tracking
One thing that comes up repeatedly on my podcast is conversion tracking.
I would say the majority of guests I've spoken with have raised some form of problem with it.
We now have incredible automation. We have AI tools. We have increasingly sophisticated campaign technology.
And yet people are still getting conversion tracking wrong.
That tells us something.
Technology doesn't eliminate the need to understand what we're measuring.
A human still needs to determine which actions matter. A human needs to understand the customer's path to purchase. A human needs to notice when something changes elsewhere in the technology stack and suddenly the data being collected no longer represents reality.
One of the oldest pieces of advice in PPC still applies:
Don't set it and forget it.
In fact, AI may mean we have more things to check, not fewer.
It can reduce the time required to perform certain tasks, but that creates another responsibility: reviewing what the automation is doing.
If you're using AI-generated insights, you still own them
This is particularly important when it comes to reporting.
Imagine you're about to walk into a client meeting with a deck containing AI-generated insights and you don't feel confident about what you're presenting.
My response is simple:
Do the work until you are confident.
Because if you walk into that room unsure about the information on the screen, the client will see it.
AI isn't accountable for that presentation.
You are.
I use AI in my own writing. But the process isn't “AI, then publish.”
It's me, AI, then me again.
I review the work. I move paragraphs. I remove things that don't make sense. I change language. I question the output. I edit.
I'm the final editor.
The same should apply to PPC.
One useful test is to present your work to somebody outside the industry. Find a friend or family member who isn't a digital marketer and talk them through it.
Does it make sense?
Does it sound human?
Can they understand the argument?
If not, you probably aren't finished.
AI-generated creative has a trust problem
Creative is another area where I think marketers need to be careful.
When I started running PPC Live events, one of the most important investments for me was photography and videography.
Why?
Because when I promote the next event, I want to show real people having a great time at a real PPC Live event.
I don't want to generate an image of imaginary people supposedly enjoying it.
If you're telling me your event is fantastic but the image you're using to prove it is AI-generated, why should I believe you?
We're now entering an interesting stage where consumers increasingly look at an image and ask a question before they've even processed the advertising message:
Is this AI?
That is an additional stage in the path to purchase that marketers need to think about.
Before someone considers whether your product or event is right for them, they may first be deciding whether the thing they're looking at is real.
That has implications for trust.
I'm not saying marketers should never use AI-generated creative. I'm saying we need to understand how audiences respond to it and make deliberate decisions rather than using it simply because it's cheap and available.
## Junior marketers still need the fundamentals
This is probably one of my biggest concerns.
People entering PPC today have access to tools that can do things we couldn't have imagined when I entered the industry.
That's exciting.
But what happens if somebody learns the automation without learning the fundamentals?
How do they know when the automation is wrong?
We need to encourage junior marketers to remain curious, to read, to understand the history of the discipline and to learn why things work rather than simply learning which button to press.
That doesn't mean rejecting AI.
It means combining the new tools with foundational knowledge.
The responsibility also sits with those of us who have been in the industry longer. We should be recommending books, sharing knowledge and explaining where today's technology came from.
There needs to be an exchange between generations.
Younger marketers can show us new tools and new ways of working.
We can help provide some of the context that makes those tools useful.
What separates a good agency from a great one?
If every agency has access to the same AI tools, the tool itself cannot be the differentiator.
The difference is what you do with it.
A great agency takes the technology and adapts it to the client.
What does this particular business actually need?
What information is useful?
What is irrelevant?
What should we automate?
What shouldn't we automate?
Think about dashboards. You wouldn't necessarily create one dashboard and give exactly the same version to every client. Different businesses care about different things.
AI should be treated similarly.
The agencies that succeed will be the ones that combine technology with context and expertise.
They won't just adopt AI.
They'll shape it.
Be cautiously curious
Curiosity is one of the most important habits anyone in PPC can develop.
But I would add another word to it:
**Be cautiously curious.**
Our industry is full of shiny new things.
Every week there seems to be another feature, another update, another tool and another LinkedIn post from somebody achieving incredible results.
Explore them.
Test them.
Learn about them.
But don't allow yourself to be dragged in every direction by hype.
Come back to the business strategy.
Come back to the client's goals.
Come back to the customer.
Those things should determine what you test and implement.
Not somebody else's screenshot on LinkedIn.
AI is still immature
If there is one thing I want PPC marketers to remember about AI today, it's this:
AI is not as mature as we think it is.
We're impressed by what it can already do, but this technology is still developing incredibly quickly.
That means we need to be thoughtful about how much responsibility we hand over to it.
We should also be thinking beyond productivity.
AI has physical infrastructure behind it. Data centres consume energy, water and other resources. As adoption increases, we need to be willing to ask questions about the long-term consequences of that growth rather than simply assuming that more AI is always better.
Technology will continue changing.
The tools we are excited about today may look primitive in ten years.
But the fundamentals of good marketing are much harder to automate: curiosity, judgment, communication, understanding people and knowing what the business is actually trying to achieve.
AI can make us faster.
It can help us analyse.
It can give us new possibilities.
But it still needs somebody steering it.
And that person needs to know where they're going.
Comments