4 arguments worth recording.

I think this is selection bias dressed up as platform performance — the brands still spending on Facebook in 2026 are the ones who already have their creative and targeting dialed, so of course their CTR is up.

Or the auction actually got less competitive because a chunk of the market moved budget elsewhere, which means the sixty-cent CPC is real and you should be buying it before everyone notices.

My read is that disclosure kills the one advantage AI creative had — speed to test at volume — because now every asset needs a legal review before it goes live.

The disclosure requirement only matters if users care, and so far there's no data showing that a label on an ad changes CTR or conversion, so this is theater until someone runs a holdout test.

I think this flips the entire zero-click playbook — if you can't control the source that the AI quotes, then optimizing your own content is just hoping the model picks you, which is not a strategy.

The experiments don't say *how much* more third-party mentions matter, and they don't separate branded search from category search, so we have a directional signal with no denominator.

My position is that if the inventory is only valuable because it's underpriced, then you're not testing a channel, you're timing a market inefficiency, and that's fine but don't build attribution models around it.

Cheap inventory with low competition is exactly when you should test, because that's when you learn whether the intent exists at all, and if it converts now it'll convert later when the auction fills in.