You made it to the final Sunday column of the year.
You didn’t unsubscribe. You didn’t rage-close the tab. You stayed through the whole thing like someone politely enduring a long meal with distant relatives.

That alone deserves a quiet nod.

Sunday columns aren’t for hot takes or bloodletting.

They’re for sitting back, stretching a bit, and admitting what stuck. Not what we said we learned on panels or LinkedIn threads. What actually lodged itself in the collective brain.

So let’s talk about that.

2025 wasn’t the year AI arrived. It was the year AI stopped asking permission.

Media buying slid from “AI-assisted” to “AI-assumed.” Performance bundles, auto-optimized video, default-on recommendations. Fewer levers, more trust. Or at least more resignation.

And with that came the collision everyone felt but few wanted to linger on:

Automation wins efficiency.
Opacity wins power.

That tension defined the year.

Everyone spent the year talking about explainability.
Everyone kept spending into systems that can’t tell you what they actually did.

Platforms reported strong ROAS while revealing less about where ads ran, why decisions were made, or what was excluded along the way. Long-standing complaints about black boxes didn’t disappear. They just got better lighting and cleaner UI.

The uncomfortable truth buyers absorbed:

If the number goes up, curiosity goes down.

Most teams knew they were letting vendors grade their own homework. Most teams also knew the homework was due Friday and the CMO only reads the summary slide.

So the industry shrugged and called it progress.

Attribution finally cracked.

Cookies faded. Privacy rules shifted mid-stream. Chrome teased, delayed, teased again. The old “clean funnel” fantasy collapsed in public.

What replaced it wasn’t elegance. It was bricolage.

MMM made a comeback. Clean rooms became normal. Attention metrics stopped being fringe. Multi-ID strategies turned into duct tape holding real budgets together.

More important shift though:

Marketers admitted that over-optimizing outcomes breaks the system.

Short-term ROAS cannibalizes reach. Last-click eats tomorrow. Incrementality matters again, not because it’s trendy, but because the alternative visibly stopped working.

No parade for this realization. Just quieter decks and fewer promises.

By late 2025, streaming wasn’t the future. It was the room everyone was already sitting in.

CTV passed linear on meaningful viewing behaviors. FAST matured from “free content” into actual programmed environments. Digital video absorbed the bulk of new brand spend. Retail-media CTV grew faster than almost anything else in the budget.

And yet complaints persisted.

Under-performance on CTV usually wasn’t structural. It was self-inflicted.

Too much targeting. Too much performance framing. Too little respect for context, content, and patience.

Buyers finally got the attention they wanted and treated it like a banner.

Then wrote thoughtful posts asking why scale felt elusive.

Retail media stopped being a channel and became a gravity well.

First-party commerce signals worked. Closed-loop measurement soothed nervous finance teams. Global spend pushed toward traditional TV levels with growth north of 20%.

But the real unlock in 2025 wasn’t search-style retail ads. It was video.

Amazon, Walmart, and others wired commerce data directly into streaming environments. Brand budgets, shoppable formats, and incrementality measurement started living in the same loop.

For the first time, attention, transaction, and attribution stopped arguing with each other and sat at the same table.

That changed budget politics more than any keynote.

By the end of 2025, people finally said the quiet part out loud:

Transparency is mostly theater.

More automation didn’t clarify spend paths. It obscured them further. Simplified buying often meant fewer questions could be asked, not better answers.

So marketers shifted focus. Less talk about openness. More talk about contracts, data rights, auditability, and independent verification.

Deeper realization underneath it all:

Fraud detection, brand safety, and AI-driven curation now decide who gets paid.

These systems don’t just filter ads. They shape the economic map of the web. Entire publishers rise or vanish based on models few people can audit and no one fully governs.

That’s not a future problem. That’s a current one.

What’s funny, in a tired way, is that 2025 proved the industry knows all this.

Everyone mocked the behaviors a decade ago. Everyone recognizes them now wearing AI branding and a shoppable QR code.

Explainable AI posts went up. Black-box budgets went up faster.

Governance decks circulated. Outcome-bundled deals got signed.

Streaming won eyeballs. Linear kept dollars. Everyone agreed it made no sense and then showed up to the upfronts like it was a sacred pilgrimage.

2025 didn’t fix advertising. It clarified it.

AI didn’t remove power dynamics. It concentrated them.
Automation didn’t end old habits. It accelerated them.
Better tools didn’t save bad strategy. They made its consequences arrive faster.

And yet, quietly, something shifted.

People are asking better questions now. About control. About incentives. About who actually benefits when systems “optimize.”

That’s not a revolution.
It’s a pause.
Sometimes that’s enough to matter.

Stay Bold, Stay Curious, and Know More than You Did Yesterday.