The Tide of AI Marketing Is Quietly Turning

A while ago, I grabbed dinner with a friend who's been a marketing director for ten years.

He was holding his beer, and suddenly asked me:

"So tell me — three years from now, is the pile of work I'm sitting on still going to exist?"

I froze for a second.

Not because the question was hard to answer. Because it was the fifth person this year who'd said exactly that to me.

AI has shoved marketing, as a profession, into a deeply awkward position. It's not that there's nothing to do — it's that there's too much, changing too fast. You learn one trick, and by the next day it's already worthless.

I recently spent a full half-year going through the topics this industry cares about most. And only after finishing did I realize —

The tide really has turned.

The Tools Haven't Changed. The Underlying Logic Has.

Let me ask you something first.

When you do SEO day to day, what are you optimizing for?

Keywords. Backlinks. Page structure. You've been doing these things for five years, ten years.

But the way users look for things has changed.

They don't open a browser and type in keywords anymore. They just ask ChatGPT, ask Perplexity, ask Doubao (ByteDance's AI assistant): "My three-year-old won't eat, what should I do?"

You see — that's not a keyword. That's a conversation.

Search is shifting from a game of "keyword matching" into a conversation about "brand authority."

Whoever answers well, whoever gets cited more, wins. How many keywords you've stuffed into your page matters less and less.

What does that mean?

It means the SEO traffic moat you spent the past decade building is slowly starting to fail. And the new moat — whether other people are willing to mention you inside the conversation — you haven't even started digging yet.

A Word That Keeps You Up at Night: "Zero-Click"

This next change hits even closer to home.

What's zero-click?

It's when a user lands on the search results page, glances at the summary, gets their question answered, and never clicks through to your site at all.

Think about it. You spend three months writing a ten-thousand-word deep dive, and the reader walks away after reading an eighty-word AI summary.

Traffic. Gone.

So is your website still worth maintaining?

Yes. But the logic has flipped.

In the past, the website was the "traffic entrance." Now, the website has to become a "trust anchor": whether AI will put you on its citation list depends on how cleanly your content is governed and how solid your brand is.

Content governance, in the zero-click era, is more valuable than ever.

Because you're not fighting for clicks anymore. You're fighting for the right to get a mention from AI.

The Real Crisis Is Hiding Inside Your Team

Now that we've talked about traffic, let's talk about people.

This is the section that was hardest for me to swallow.

There was a survey that asked marketers: are you satisfied with your current job?

Most of them said, yes, satisfied.

Then came another question: in the past six months, have you been looking at new opportunities?

More than half said yes.

You see — "satisfied" and "getting ready to leave" can be two states of the exact same person.

Why? Because AI has stripped the sense of security out of every single role. Copywriters worry about being replaced by text generation models, designers by image generation models, market analysts by AI agents. On the surface they're still doing the work, but inside they're already quietly updating their resumes.

What's worse is that companies are also churning their workforce: using AI to replace some junior roles, while the people they're hiring all understand prompts and workflows.

The veterans want to leave, the newcomers are hard to hire, and a gap is opening up in the middle.

Some people call this the "ghost workforce." The work is still getting done, but the "role" doing the work — nobody can say anymore whether it's human or machine.

Can AI Make Decisions for You?

For this one, I have to tell a story.

There was a CMO who wanted to test whether AI could do market research. He fed the model the target audience, the survey draft, all the historical data, and asked it to generate a report.

And the result?

The report looked professional, the charts were beautiful, the conclusions sounded reasonable.

But when he checked it against real user interviews — half the conclusions were wrong.

Where did it go wrong? It went wrong because AI treated "the statistically most likely answer" as "the real answer."

It doesn't know that inside your user base there's a group of people who "say yes to your face while cursing you under their breath." It doesn't know that in this industry, the real decision-making chain plays out over dinner, not inside a survey.

What does that mean?

AI is a fantastic copilot, but if you let it be the captain, it's going to crash.

The CMO's job isn't going to disappear. But it's being sliced in half: one half is what AI can do — data wrangling, first drafts, competitor scanning; the other half is what AI can't do — judgment, negotiation, dealing with human nature.

You'd better figure out which half you're training for.

A New Phrase: "Agentic Workflow"

This last one, I think, is the most worth chewing on this year.

What's agentic workflow?

Put simply — you don't direct AI to do one task, you let AI run an entire workflow by itself.

For example.

In the past, when you used ChatGPT, you'd write a prompt, and it would return a passage. The two of you went back and forth, like playing ping-pong.

Agentic workflow is different. You give it a goal: "Help me monitor all the negative brand mentions across the entire web this week, tier them, summarize them, draft response talking points, and send it to me Friday afternoon."

And then it goes off on its own — pulls the data, runs the classification, drafts the work, and pushes the finished product to you right on time Friday.

From "tool" to "employee."

But there's a huge trap here.

Whether an agentic workflow runs well is 90% determined by whether the content you feed it is clean.

If your content assets are a mess — outdated information, mutually contradictory documents, knowledge bases nobody maintains — the agent will treat that garbage as truth and amplify it back out, multiplied.

So looking back, the core anxiety of this entire industry in 2026 can actually be boiled down to a single sentence:

AI hasn't come to help you do more things. It's come to amplify everything about you — including your chaos.

To Wrap Up

Writing this, I'm reminded of the question my friend asked me.

"Is the pile of work I'm sitting on still going to exist three years from now?"

My answer is: the work is still there, but the way of doing it has completely changed.

The traffic entrance has switched, the content moat has shifted, the sense of job security is gone, and even whether you can trust AI to make decisions has become a real question.

But there's one thing I'm actually more certain of than I was half a year ago —

People who can exercise judgment will never be replaced.

AI can write, calculate, scan data — but it doesn't know that when a client says "let me think about it" over dinner, that's either politeness, or they genuinely didn't buy what you were selling.

That kind of thing — only you know.

So don't panic about the tools. Build your judgment first.

That's a reminder to myself as much as to anyone.