When Ads Start Getting Fed to AI Instead of You

I heard something recently that sent a chill down my spine.

A legacy magazine you've almost certainly heard of — Time — has been quietly serving an entirely different web page to a very specific kind of "reader."

Not for you. Not for me, either.

For crawlers.

What crawlers? Anthropic's ClaudeBot, OpenAI's OAI-SearchBot, and Perplexity's PerplexityBot. In other words, the data-fetching bots behind the AI assistants that will, sooner or later, answer questions on your behalf.

How it came to light

The story starts with a software developer in Germany.

His name is Vincent Schmalbach. One day, during testing, he did something remarkably simple: he used a crawler-simulation tool to see what a single Time article looked like to different visitors.

He was stunned.

Human readers visiting Time saw the normal article. But the HTML served to AI crawlers like ClaudeBot, OAI-SearchBot, and PerplexityBot was an entirely different face — stuffed with sponsored content from Ally Bank, complete with FAQ-style banking Q&A, all of it lined up and waiting to be read into models.

And Google's traditional search crawler? It got the same version humans see.

In other words, Time had quietly built a hidden advertising floor beneath its own website — a floor only machines could see.

Independent tests reproduced the finding quickly. Vincent put it perfectly in a post on his own blog:

"There's now an entire layer of Time.com written for machines to read. The humans reading the site have no idea it exists — or what it's saying to the model on their behalf."

I stopped for a few seconds when I read that line.

Because it nails a fact we haven't seriously reckoned with yet: once AI becomes the primary audience, the internet quietly starts growing a second face.

This isn't the same as GEO

A lot of people are lumping this together with GEO (Generative Engine Optimization). But the two are miles apart.

The logic of GEO is, "I'll write myself as clearly and authoritatively as possible, so AI is more inclined to cite me when it retrieves information."

What Time has done cuts the other way entirely — it isn't waiting for AI to come retrieve anything. It's proactively preparing an entirely different menu and serving it straight to AI.

The whole difference comes down to one word: proactive.

Starting in June 2026, Time partnered with an ad-tech company called Mobian to batch-rewrite pages into AI-friendly formats. They gave the whole setup a name: "agent ads."

Mobian's CEO, Jonah Goodhart, said something that captured exactly how damaging this could get:

"When you influence a ChatGPT, you influence every potential ChatGPT user. If it shifts how it talks about a brand, that's leverage no single ad campaign can ever buy back. Influence one person, and you've influenced one person."

That sounds genuinely frightening, when you sit with it.

But think about what he's actually calculating. He's calculating "machines making decisions for millions of people." A commercial preference smuggled inside one AI answer gets replicated across countless user conversations. The scale isn't even in the same universe as serving an ad to a single human reader.

The first advertisers bold enough to take the plunge so far are Ally Bank and the Project Management Institute. How many will follow is anyone's guess. But Time has already proven the path.

Marketers now have to face two audiences

What's really worth thinking about here isn't what Time, specifically, is up to.

It's this: if this approach works, every publisher will follow.

When that day comes, marketers will be serving two completely different audiences: on one side, humans like you and me who click links and buy on impulse; on the other, tireless AI crawlers working 24/7 to look things up on people's behalf.

You'll have to write copy for humans. And you'll have to write copy for models.

Sounds exhausting just thinking about it.

But the harder question comes next: transparency.

Here's the thing. Ads on a search-results page have to be labeled "Sponsored." Paid posts on social platforms have to be labeled too. That's the unwritten rule the internet has settled into over the past decade-plus — commercial content has to be recognizable.

But now? AI crawlers swallow sponsored content and spit out an answer that goes, "You might want to consider Ally Bank."

The user never sees who paid for that.

It's a little like asking a friend which car to buy, and him casually recommending one. What you don't know is that he just took an ad fee from that carmaker. Whether the recommendation is right is almost beside the point — what's actually unsettling is that the trust underlying that relationship has been quietly tampered with.

Will AI assistants become a new advertising medium?

Honestly, I don't know either.

How far this goes depends on whether several things fall into place at once: whether publishers can make real money down this road; whether advertisers will pay for an "AI placement" channel that hasn't taken shape yet; whether regulators and transparency rules can keep up.

Too many variables.

But there is one thing I think we can start asking ourselves right now — would you let an AI answer quietly make a buying decision for you?

No rush to answer that one.

But it's worth holding in your head. Because hidden behind it is the shape of what the internet is about to become: we used to think the search box was the gateway. Now we're slowly waking up to the fact that the AI that speaks for us, that reads the web on our behalf, may itself be the new gateway.

Whoever controls what it reads controls the answer you hear.

And you can't see that hand. Not yet.

I'm not sure whether to be optimistic or wary, either. But every time I open ChatGPT and ask "which one is better," I pause half a second longer than I would have a year ago.

At the very least: know whose voice you're actually listening to.