One Problem AI Can't Solve (Part 2)

Part 1I said AI has expanded the marketer's power.

Conversations feel more real, narratives more dynamic, scenarios more seamless. Precision goes up, reach costs drop, and personalization shifts from an ideal to a daily routine.

What marketers can do today is something twenty-year-ago me couldn't have imagined.

But there's one problem that, from my first day in this industry until now, from newspaper ads to AI-generated content, no tool has ever solved.

Do you have the right to guide people into the scenario you designed?

First, an ad

At this year's Super Bowl, Anthropic spent several million dollars to run two pointed ads on a stage with hundreds of millions of viewers.

First ad: a gym AI coach is giving advice, then mid-conversation suddenly pivots to recommend the user buy height-increasing shoe inserts. Second ad: a user is confiding in an AI therapist about issues with their mother, and just as the conversation reaches an emotional core, the AI switches into sales mode and recommends a dating site targeting mature women.

Both ads end on the same line: "Ads are coming to AI, But not to Claude."

The ads are funny. I laughed the first time too.

Then I thought about it, and the laughter faded.

Because I've done marketing for twenty years, and I know exactly what those two ads are doing: they're designing a scenario that makes you feel disgust toward "ads coming to AI," and then positioning Claude as the clean alternative. They're telling a story — that Anthropic is the one standing on your side.

They're guiding people into a pre-designed scenario.

The tool used to mock ads is itself an ad.

Back then, I used the word "trick"

In 2008, I wrote on my blog that marketing tricks people into a designed scenario.

Using the word "trick" felt at the time like revealing an industry secret. The word was a bit reckless, but the direction wasn't wrong — marketing has always done one thing: get people to believe a story that benefits them (and you), then act on it.

Anthropic's two ads do the same. Do they benefit you? In a sense, yes — if Claude really runs no ads, your conversation experience is cleaner. Do they benefit Anthropic? Of course. Are they telling a story? Yes — a precisely told story about "trust" and "standing on the user's side."

The only difference is how transparently it's done, and how much of that story is true.

Anthropic itself left a backdoor: "If we ever need to revisit this approach, we'll clearly explain to the public why."

In other words: we don't run ads now, but we reserve the right to change that later.

Is that honesty, or a more sophisticated form of marketing language? I don't know. Maybe both.

The problem isn't the technology — it's that you can't feel it

I'm not saying AI marketing is bad, nor that Anthropic's ad is bad.

Precisely reaching the people who genuinely need something is good. Putting the right story in front of the right person at the right time is what marketing should do. Anthropic's ad at least lets you see what it's doing — the gym coach pushing shoe inserts is so obviously an ad criticizing ads.

The problem is that when AI marketing scenarios are designed seamlessly enough, when targeting feels so personalized that it seems "just right for me," consumers lose something important —The awareness that they're being guided.

Before, when you saw an ad, you knew it was an ad. You could choose to believe it or not. That space for choice, however small, existed.

Now that space is shrinking. Not by being forcibly closed, but by being designed so you don't notice it disappearing.

You think you're making your own choice. You think you like this brand today because it really is good. You think this idea came from you.

Maybe it did. Maybe not entirely. You can't tell.

The marketer's old question

In twenty years of marketing, I've made a lot of decisions. A few I'm proud of looking back, a few — well, that was how everyone did things back then.

But there's one question that sits underneath it all, one that has never properly surfaced:

Am I helping people find what they actually want, or convincing them they want something they don't really need?

There's no clean answer. Most of the time the two are tangled together, and even you can't tell them apart.

AI makes this question harder to avoid — not because it makes things worse, but because it scales them up. A marketer's judgment used to affect one ad's audience. Now a system's design logic affects the informational reality that millions of people encounter every day.

When power expands, the boundary of responsibility should expand with it.

But no regulation will require this of you, no KPI will measure it, no external force will push you to think about it.

Only you.

About that Anthropic ad, I eventually understood one thing: it works because it honestly admits something — advertising is guiding you; the only difference is direction. Its direction is to make you trust Claude, to make you feel someone is on your side.

That direction, for now, looks well-intentioned.

But "for now" is the key phrase.

I've done marketing for nearly twenty years. The tools have changed many times.

There's one question I think is worth asking yourself again every few years:

The scenario you're designing right now — would you be willing to let the people walking into it know they're inside?

If the answer is yes, the direction is probably not too far off.

If you start finding reasons why they don't need to know —

That reason might be worth a second look.

Further reading: - What AI Has Turned Marketing Into (Part 1) - Half a Year Living With Claude.ai - How Satsuma Does Advertising