AI Is Here — Should Marketers Be Excited, or Afraid?

The other day, I grabbed dinner with a friend who works in branding.

He picked up a piece of tripe with his chopsticks and, out of nowhere, asked me: "Tell me — if I let AI write my ad copy from now on, do I never need to hire a copywriter again?"

I said, "Put the tripe down first. We need to have a real talk about this."

Because the answer is far more complicated than you think.

First, a fact: the tools have changed, and so have the rules

This wave of AI, for anyone in marketing, is an opportunity.

I don't doubt that for a second.

Think about it. Writing an email, putting together a report, editing a short video — that used to take three to five people a full week. And now? One person, one chat window, half a day, done. Efficiency has been completely rewritten.

But.

There's always a "but."

The tools are moving faster than the rules. What you're allowed to do, what you're not, what'll cost you money, what'll blow up in your face — these "traffic rules" are still being paved while the cars are already driving on them. You floor it, and there might be a sinkhole ahead.

So I told my friend: don't rush to fire anyone. Let me first walk you through a few hurdles every marketer is going to have to clear from here on out.

Hurdle 1: Can you tell what's actually real?

AI has a problem when it writes things.

It can fabricate with a perfectly straight face.

Ask it to write a product description, and it'll bring a feature that doesn't exist to life in vivid detail. Ask it for an industry analysis, and it'll invent a string of numbers that look impressively professional — with sources it made up entirely.

The industry term for this is "hallucination" (when an AI model confidently outputs false information it generated itself).

What does hallucination mean? It means the model doesn't even realize it's making things up. It's just calculating probabilities, guessing which word should come next. And as it keeps guessing, it assembles a fake story that looks remarkably real.

Who's responsible for that?

You.

That's right — you, the marketer who hit Enter. The model isn't on the hook. The platform isn't on the hook. The finger-pointing goes round and round, and the blame ultimately lands on whoever wrote the prompt.

That's why any company of even modest size is now building so-called "AI governance" systems. In plain terms: putting a cage around the AI to stop it from outputting biased, inaccurate, or outright fabricated content.

But as we all know, no cage is ever foolproof.

My friend's chopsticks froze when he heard this.

So what do we do?

I said: at the very least, there's one thing you have to insist on — make the AI cite its sources, plainly and completely. Which sentence came from where. Where each number originated. Only then can you actually use its output, and only then can you trace things back if something goes wrong.

Hurdle 2: Who does this thing actually belong to?

This is the question giving legal departments migraines right now.

You use AI to generate a poster. Who owns the copyright — you? The AI company? Or the thousands of original creators whose work was used to train the model, most of whom probably have no idea they're even in there?

Nobody can say for certain.

Take it a step further. You ask AI to design a new product concept from a single line of description. Is that concept your idea, or the model's?

Worse still: the model grew up eating data from across the internet. If it "spits out" a chunk of copyrighted content verbatim and slips it into your ad — you might be infringing someone's rights without even knowing it.

The lawyers are already gearing up.

I told my friend: this year, you absolutely have to sit down with legal for at least one walkthrough on this. Don't wait until something blows up to start asking questions.

Hurdle 3: Will your brand get "smoothed flat" by AI?

This is what truly worries me.

Think about how AI works. It learns from what already exists. It chews up all the copy, images, and video on the internet, digests it, and spits out the version that looks "most presentable."

But "most presentable" usually means "most average."

It won't hand you that weird, electrifying idea. It won't deliver that tone and cadence that only your brand has. It'll give you something you can't find fault with — but that no one will remember either.

And honestly, that's the scariest part.

People who've spent twenty years managing brands are most afraid of exactly this. The brand voice you painstakingly built, that distinctive, instantly recognizable "this is you" quality — after one pass through AI, it turns into generic, interchangeable stuff that could belong to anyone.

Twenty years ago, the internet already knocked brand management off target once. Everyone started staring at clicks, conversions, follower counts, and forgot all about the "long-term brand equity" North Star.

Now AI is here, and that force is even stronger.

Making content is going to get easier and easier. Making content that is unmistakably your brand is going to get harder and harder.

When my friend heard that line, he set his chopsticks down.

Hurdle 4: Fake ads are drowning the internet

Let me spend a little more time on this one, because it's worse than you'd think.

There's a category of company that uses AI to mass-produce SEO content and ads at scale. The business model is dead simple: for every 1 unit you spend on ads, you earn 1.25 back. Sounds small, right? But they push out billions of pieces.

Billions.

The result: legitimate companies' content gets buried under this AI garbage, and users never scroll far enough to see it.

And then there's the nastier stuff. Someone made a fake video of a well-known UK consumer-rights figure endorsing a certain investment product, and even dragged in a tech titan to back it up. All completely fake. The video was good enough that even the real person would have had a hard time spotting it immediately. Reported it to the platform? Vanished without a trace.

A group of researchers at the University of Zurich ran an experiment that's pretty revealing.

They had AI write a batch of fake tweets, and they also scraped a batch of fake tweets written by real people. Then they brought in a group of humans and asked them to guess: which ones were real, which were made up.

The result? People were more likely to believe the AI-written versions.

Why? Because AI writes more coherently and reads more smoothly. And humans, by nature, are easily persuaded by things that "look logical."

And that advantage is only going to grow.

In other words, this blade of AI is pushing information further from the truth, not closer to it.

Hurdle 5: Bias, hiding in the deepest layers of the data

What did AI grow up eating? Data.

Who produces that data? People.

What do people have? Bias.

So AI inheriting bias is essentially inevitable. It might, without even realizing it, amplify stereotypes about certain groups, or discriminate against certain users inside its targeting logic.

Can this be fixed at the root? Hard. But marketers should at least fight for one thing: for every dataset they use, know where it came from and what algorithm processed it.

You need a bird's-eye view, watching over this data-production assembly line.

Hurdle 6: No matter how powerful the tool, that last 20% is on us humans

This one is where my friend and I, by the end of our conversation, were both nodding.

A global CMO (Chief Marketing Officer) from Deloitte put it beautifully. She said: the machines do that 80% — processing massive datasets, running endless variants, grinding through all the exhausting, repetitive work. Humans do the final 20%.

That 20% is emotion, empathy, strategy, and the kind of inspiration only a human can generate.

Humans are emotional creatures.

Marketing, at its core, is about stirring emotion. And no matter how many more versions AI iterates through, it is not human. It can write poetry that reads more and more like a person wrote it — but it will never be moved by a poem itself.

And that, precisely, is where the boundary lies.

A closing thought

When my friend and I finished dinner and went to pay the bill, he asked one more thing: "So — do you think I should use it or not?"

I said: You should. You have to.

But use it with your head.

Set the rules clearly. Bring legal in. Hold the line on your brand. Stay locked onto "what's actually real." Be optimistic where optimism is warranted, and be on guard where caution is.

You need both mindsets. Neither one is optional.

Because this wave of AI isn't waiting for anyone. It's already rewriting the marketing industry. If you don't get in the game, your competitors will.

But remember one thing —

The tool is the tool. You are you.

The one who decides the soul of the content is never the model — it's the person who presses Enter.

That night, when I got home, I wrote the conversation down. Because I realized this isn't just my friend's confusion alone.

This is probably what every person in marketing is thinking about this year.

Once you've thought it through, stop hesitating. Get in the game.