Marketing engineers build the machine. Someone still has to edit it.

The latest marketing job title to attract attention is the marketing engineer. A Wall Street Journal article describes the role as part builder, part marketer: someone who can work with AI agents and automation but still bring taste, judgement and an understanding of what good marketing looks like.

I can see why the title is catching on. AI is making the mechanics of marketing more programmable. Competitor research can be automated, paid media systems connected and personalised creative generated at scale. AI-search monitoring can become a repeatable process, and reporting can be assembled without somebody manually moving information between tools every week. A marketer who can design and improve those systems is doing something meaningfully different from somebody who only operates inside them.

The discussion becomes more interesting where the engineering metaphor starts to run out. The more marketing becomes automated, the more organisations need somebody to decide what the machinery should pay attention to, which evidence deserves to be trusted, what should be published and when the system is confidently producing the wrong thing.

The marketing engineer still needs an editor.

I have already asked whether I am a content engineer

A month ago I published Am I a content engineer?. I had noticed that my own work was moving away from individual outputs and towards the systems around them: evidence capture, reusable customer intelligence, GEO inputs, knowledge repositories, AI routing and workflows designed to keep useful context after a project ends.

I could recognise myself in the newer definition of content engineering, but I was uncomfortable with the title. Part of the reason was straightforward. I am not a software engineer, and I do not want to borrow the authority of another profession simply because AI has made technical language fashionable.

The more important reason was that “engineering” did not quite describe the responsibility I cared about most. I was not primarily trying to make content production more automated. I was trying to make editorial decisions more informed, recoverable and reusable. The WSJ piece makes that tension much clearer.

The engineering part is real

There is an easy response to every new marketing title: call it rebranding and move on. Marketing has had growth hackers, ninjas, evangelists, storytellers and plenty of other labels that promised a new discipline while often describing familiar work in fashionable language.

The WSJ article raises that criticism directly. Some marketers have been automating tasks, connecting systems and building workflows for years, and adding “engineer” does not automatically make the work more rigorous. But I would not dismiss the underlying shift, because the work itself is changing.

A marketer can now build an agent that monitors a category, compares competitor claims, prepares a weekly briefing and routes important changes to the right person. A content team can connect transcripts, research, customer evidence and previous editorial decisions so that a new brief starts with accumulated context rather than a blank document. A GEO team can sample AI answers, identify recurring sources and use the result to decide what evidence needs to be strengthened. Those are systems problems as much as content problems.

The useful part of “marketing engineer” is therefore not the prestige of a technical title. It is the recognition that marketers increasingly need to understand how information moves through a system. What enters the workflow? What is transformed, retained or automated? Where does a human intervene, and what happens when the output is wrong?

But marketing is not a deterministic system

The engineering metaphor gets weaker when we move from workflow to judgement. In the WSJ article, Profound co-founder James Cadwallader describes the human as the person who must orchestrate and adjudicate what the agents should build. He also describes marketing as “non-verifiable by nature”.

That is an important qualification. A workflow can be engineered, and so can a data pipeline. An agent can be instructed to collect competitor pricing, identify changes and send an alert. But many of the most consequential marketing decisions are not verifiable in the same way.

Is this customer story actually persuasive? Does this research finding support the claim we want to make? Is this a useful connection between two weak signals or merely a coincidence? Should an organisation respond to a popular search question, or is the premise of the question itself misleading? Does a piece of evidence deserve more weight because it is repeated, or less because all the sources are copying one another?

Those are editorial questions. They require interpretation, evidence, context and judgement.

AI makes the editor more important, not less

There is a strange consequence to making production cheaper. When a team can generate ten campaign concepts instead of two, the scarce resource is no longer the first draft; it is deciding which concept deserves to survive. When an agent can monitor hundreds of sources, the scarce resource is not collection but deciding which change is meaningful. When a model can rewrite a message for twenty audiences, the difficult question becomes whether the original claim was worth making in the first place.

This is why I keep returning to editorial judgement. I don’t mean copy-editing at the end of a production line, or correcting punctuation after the machine has done the “real” work. Editorial judgement moves upstream. It decides which questions deserve investigation, what evidence counts, where claims are too weak, which contradictions need resolving and when an apparently efficient workflow is optimising the wrong thing.

The more capable the machinery becomes, the more expensive bad judgement can become, because the system can scale it so quickly.

The marketing engineer and the editor are not opposites

This is not an argument for protecting an old editorial profession from technical people. The interesting roles are likely to combine capabilities that used to sit apart.

The marketing engineer needs enough marketing judgement to know what the system should optimise for. The editor increasingly needs enough technical fluency to understand what can be automated, what context an AI system needs and how knowledge should be structured so that it can be reused. That is close to the territory I have been calling Editorial Intelligence. The point is not to turn editors into developers; it is to move editorial judgement into the architecture of the system.

If a customer interview produces useful evidence, how is that evidence retained after the case study is published? If a conference produces three strong observations, how are they compared with previous thinking instead of disappearing into a recap? If AI-search monitoring exposes a weak answer, does the team automatically produce another page, or first ask whether it lacks the evidence needed to answer the question credibly? Those are design decisions, and they are also editorial decisions.

Building the system is not the same as deciding what it’s for

When I wrote Am I a content engineer?, I was mostly trying to describe the work I found myself doing. The WSJ piece gives me a clearer distinction. Content or marketing engineering is about building the system around the work. Editorial Intelligence is about the evidence, context and judgement that tell the system what deserves to happen.

The two overlap. A good system should preserve context, make evidence recoverable and route work intelligently. It should reduce repetitive production and allow people to spend more time on higher-value decisions.

But infrastructure alone does not create a point of view. An agent can retrieve every customer transcript in the company and still draw the wrong conclusion. A perfectly structured evidence repository can contain weak evidence. An automated GEO workflow can identify that competitors are appearing in AI answers without understanding why a buyer should choose one of them. The system needs somebody, or some explicit process, responsible for interpretation. That is the editorial layer.

The same distinction appears in AI-native marketing

At Semrush Spotlight, I heard a related argument from an OpenAI marketing leader: redesign the work around what current AI tools can do rather than bolting AI onto the old workflow. I wrote about that in What OpenAI’s marketing workflows reveal about the future of marketing.

The implication is bigger than automating deliverables. If reporting becomes largely automated, what decision was the report supposed to support? If creative variants become almost free, who decides which audience insight matters? If an agent can recover organisational context before a brief is written, what new questions should that context help surface?

That last question matters to me. In my recent Event Intelligence experiment, the useful result was not simply that AI helped me turn conference notes into articles. The system connected a Semrush session on YouTube with existing GEO work, customer video and evidence thinking, and surfaced an opportunity I had not consciously identified: YouTube itself should be treated as part of a GEO evidence strategy. That was not production automation. It was contextual judgement assisted by a system.

So what should we call the person?

Maybe marketing engineer will stick. The WSJ reports that companies are already hiring for the role and measuring success through deployed agents, time saved and commercial impact. For some organisations, that will be exactly the capability they need.

But I would be wary of treating “engineer” as the destination for every marketer who becomes technically fluent. The title risks implying that the highest-value work is building the machine. Sometimes it will be. Sometimes the more valuable contribution will be deciding what the machine should know, what it should ignore, what evidence needs to be collected before it acts and which output should never have been generated.

I still would not call myself a content engineer. I am much more comfortable saying that I do some content engineering while working on a broader problem: how editorial judgement, evidence and accumulated context operate inside AI-enabled marketing systems.

The distinction feels more important now than it did when I wrote the first piece. AI is creating a need for more builders, while also making the people who can edit the system itself more valuable.