Ads Made by AI Are So Cheap They Feel Free. But What Comes Free Usually Has an Owner

A while back, an old friend invited me out for tea.

He handles business and legal affairs at a creative agency — half his time on set, half of it buried in contracts. In his own words, his job lately has started to feel like the same thing on repeat:

Talking people down.

And who? Brand teams that show up, thrilled, holding AI-generated assets and asking, "Can we run this as-is?"

There are more and more of them. Enough that over three pots of tea that afternoon, the two of us talked the whole subject through, from every angle. By the end, I was left with one feeling: the traps hiding in here are worth sharing with more people who build brands.

Because a lot of people haven't thought through one question:

When AI generates something, is it actually yours — or someone else's?

First, Let's Be Clear: What "Generative" AI Actually Means

You might say: I've been using AI forever — research, summaries, weekly reports.

True. But that kind of AI is, at heart, an assistant. You ask, it answers; you tell it to find something, it finds it for you. The work is still yours — it just makes you faster.

Generative AI is different.

It does the work itself. Images, music, video, copy — you describe it in a sentence, it delivers the finished product.

One is an assistant. The other is a ghostwriter.

An assistant helps you look things up; a ghostwriter writes the piece for you. The difference looks small. It's actually a dividing line between two eras.

What does that mean for people who build brands? Let's run the numbers.

A 15-second ad used to mean script, casting, shoot, post-production — the whole run could easily cost hundreds of thousands, with timelines measured in months. Now? One person, one computer, one afternoon, four options to choose from.

Is it fast? Fast.

Is it powerful? Powerful.

For marketing teams who complain day in and day out that the budget is never enough, this is the kind of good news they wouldn't even dare dream of.

But.

What Comes Free Usually Has an Owner

My friend said something that stayed with me. A lot of people treat generative AI as a shortcut, without ever considering:

At the end of that shortcut may be a liability.

Why? Let me walk you through three traps.

Trap One: Nobody Knows Where Its Food Came From

Why is generative AI so capable? Because it grew up fed on an ocean of existing works. Images, text, music — a fair share of it taken without so much as a word to the original creators.

The lawsuits have already begun. Stock-photo companies have sued. So have writers. All of them point at the same question: You trained yourself on my work — did I ever agree to that?

What does that have to do with brands? Think it through.

The assets you generated with one of these engines went into your ads, onto your packaging. If one day that engine's training data is ruled infringing, will the assets in your hands still hold up?

It's as if you hired a ghostwriter, only to discover he'd been memorizing passages from other people's books and writing them out as your content.

Trap Two: What Goes In Doesn't Come Back Out

The second trap is sneakier.

In quite a few countries and regions, content generated by AI currently enjoys no copyright protection at all.

What does "unprotected" mean? It means anyone can use it.

And how do brands use AI? They feed their logo, their brand colors, their trademarks — everything at once — into an engine, and ask it to generate assets "in our house style."

That's exactly where the trouble is.

Once those most valuable brand assets are inside the system, anyone who can write a prompt may well rebuild a "you." The brand recognition you spent twenty years growing becomes someone else's raw material.

What's that equivalent to?

It's equivalent to photocopying your house key ten thousand times — and then pinning them all up on the neighborhood bulletin board.

Trap Three: The Audience's Eyes Are Faster Than the Courts

Even if you dodge the first two traps, there's a third: how it looks.

AI-generated content often has an indescribable oddness to it. Wrong number of fingers. Light and shadow that don't obey logic. Copy with a whiff of machine translation about it. Each one a small flaw on its own — but pile them up, and one vague thought forms in the audience's mind:

This brand doesn't seem to be trying very hard.

No cease-and-desist letter. No trending hashtag. Just, slowly, people stop believing you.

So Can You Use It or Not?

At this point I asked him: So by that logic, we should just not use it?

He said no. This stuff is genuinely fun — for creative experiments, it's a blast.

But "fun to play with" and "ready to slap onto your brand" are two different things. His advice was down-to-earth.

Read the terms first. Every engine's licensing rules are different — who owns what gets generated, whether it can be used commercially, who bears the liability if something goes wrong. Get it all in black and white before you proceed.

Use it once, log it once. When, which engine, what was generated — keep a record of it all. If things ever go wrong, that record is what gives you a leg to stand on.

And bring in someone who knows the field to check the work. Just as finance needs audits, creative output needs someone reviewing the risk.

One more thing: don't just tally the money saved — put the long-term ledger of brand reputation on the table too.

It comes down to one line:

Treat AI as one part of your creative strategy — not a miracle cure for everything.

We're Still in the "Dial-Up" Era

As we were leaving, my friend added a bigger-picture thought. I thought he put it beautifully.

For all the noise right now, generative AI is still at a very early stage. Like what?

Like the first few years of the internet.

The bubble burst back then, and it left a mess everywhere. But did the internet die? No. It just found a new way to live, and grew new rules.

Generative AI will most likely take the same road. Hiring at creative agencies will change. How brands manage their assets will change. Because for the first time, producing content has become this cheap, this disposable, and this endlessly editable. Between human creativity and machine convenience, a new balance will get worked out.

So the question has never been "to use it or not."

The question is "how to use it."

That afternoon, the tea was refilled three times. As we parted, he said: what this thing leaves behind, in the end, will be the people who know how to use it.

It's never the tool that makes a person obsolete — it's other people who know how to use the tool.

May you be the one who figures it out first.