Manifesto

AI is the new interface.
Who controls yours?

The shift

AI is becoming the primary interface to information, decisions, and eventually action. Not in some distant future—right now. People are asking AI for medical guidance, financial options, product recommendations, legal interpretations. The answers they get shape what they do next.

And just like the web before it, this new interface is already starting to reflect the incentives of the platforms that control it.

The problem

Ads are showing up in AI conversations. Responses are being shaped by partnerships, training data, and commercial interests. But this time the stakes are different.

On the old internet, you could see a banner ad and decide to ignore it. You could view source. You could install an ad blocker and move on. The manipulation was visible, even if it was pervasive.

In AI, the influence is embedded in the answer itself. It's not about what content you see—it's about what answers you get, what options are presented to you, and what decisions you make as a result. You can't view source on a large language model.

The gap

Right now, users have almost no way to:

See when something is being influenced or targeted
Understand how outputs are being shaped
Set boundaries on what they will or won't accept
Control how AI systems behave on their behalf

We're moving from an internet where users browsed information to one where systems interpret and act on their behalf. That creates a much deeper form of dependency—and a much greater loss of visibility and control.

What we're building

adblocked.ai is not an ad blocker. Or rather, it is—but that's just the starting point.

We're building a user-side layer that brings back visibility, control, and agency in AI systems. Not owned by the platforms. Not optimized for their incentives. Aligned with the user.

Start simple:

Block ads in AI conversations
Strip obvious manipulation

Then expand:

Surface bias and commercial influence
Show how outputs are being shaped
Allow users to define boundaries and preferences
Enable users—and eventually agents—to operate under those constraints

Over time, this becomes a control and observability layer for AI.

The last version of the internet trained us to accept invisible manipulation. This version will go further—unless something exists to counter it.

Team