The Advertising Thesis

“What is advertising really about?”

This is a complicated question. Here’s my straightforward answer:

From the Latin advertere (to turn towards), advertising is about steering people towards solutions.

Not tricks.

Not persuasion.

At its best, advertising is how businesses help real people solve real problems.

It increases social surplus by making the right option discoverable at the moment of intent.

The modern failure mode

Somewhere along the way, we swapped “help people find solutions” for “extract as much data as possible and shove whatever converts.”

That’s why ads feel gross.

That’s why people install blockers.

That’s why trust is collapsing.

And it’s also why the next platform shift (LLMs, agents, conversational interfaces) can’t just copy-paste the old ad-tech stack.

The platform shift: intent becomes the inventory

In search and feeds, inventory is pages and slots.

In chat, inventory is decisions.

A good assistant doesn’t just show you ten links.

It helps you pick.

So the new ad unit is not a banner.

It’s not even a link.

It’s a recommendation, a citation, or an action.

That means the core question changes from: “Did we get a click?” to: “Did we actually help?”

Data is not the enemy. The social contract is.

The internet trained everyone to believe the problem is data.

I don’t buy that.

Data is useful.

Personalization is useful.

The problem is when data is taken, hidden, sold, and used against you.

So the real unlock is a new social contract:

• Users should know what is being collected.
• Users should control what it is used for.
• Users should be able to say “no” without the product breaking.
• Sensitive data should be protected by default (encryption, minimization, purpose limitation).
• Ads should be legible: clearly labeled, grounded, and auditable.

When you do this right, something counterintuitive happens:

People share higher signal intent.

They get better outcomes.

Trust goes up.

Conversion goes up.

What LLM-native advertising should optimize for

Old ad-tech optimizes for impressions, clicks, and conversion.

LLM-native ad-tech has to optimize for a three-way win:

1. User satisfaction: the assistant stays trustworthy.
2. Advertiser ROI: outcomes are measurable and real.
3. Publisher retention: ads don’t create churn.

If you can’t protect trust, you don’t have a product.

You have a short-term arbitrage.

Here’s the shift in a single table:

The new primitives (ad formats that actually fit the medium)

The dominant formats won’t look like ads.

They’ll look like help.

• Verified Citations
  • Sponsored sources that are explicitly labeled and grounded in real, verifiable claims.
• Native Suggestions
  • The “right next step” recommendation when the user is clearly shopping or deciding.
• Action / Lead Cards
  • Ads that are actions: sign up, book, apply, get a quote, start a trial.

In a world of agents, the best ad is the best tool.

If you want a concrete definition of “good,” it’s something like:

If sponsored:
  disclose clearly
  ground claims
  log the rationale (for audit)
  optimize for user outcome, not just conversion
Else:
  behave normally

The thesis

Advertising should be the mechanism that funds free software without poisoning the user experience.

The future of advertising isn’t better targeting hacks.

It’s not an arms race of creatives.

It’s a system that:

• understands intent,
• respects privacy,
• makes sponsorship legible,
• and optimizes for long-term trust.

If you're interested in this vision of the future of ads, I made a landing page for it: Infrarad.