Here is the sentence nobody wants to say out loud at the next IAB happy hour, somewhere between the branded cocktail napkins and the panel titled "Trust and Transparency in the AI Era":

The has slop won.

Not "is winning." Not "poses an emerging challenge for supply-side stakeholders." Won. Past tense.

The internet you buy media on is now, by several credible counts, mostly machine-generated, and the part of it that is still human is increasingly written by a person babysitting a language model.

Let me put numbers on the funeral. Ahrefs looked at 900,000 web pages created in April 2025 and found that roughly three out of four contained AI-generated content.

A quarter were what they generously called "pure human." SEO firm Graphite clocked AI-written articles crossing the fifty percent line of everything published online by around October 2025. Pinterest, of all companies, went and cited academic literature saying 57 percent of all online content is now generated by a machine, and it did so in a legal-adjacent document, which is corporate for "we checked with the lawyers and yes, we're saying the quiet part."

So when someone tells you AI slop is a coming wave, correct them. It is not coming. It arrived, unpacked, and changed the locks.

Now here is the part that should make everyone who spent 2023 patting themselves on the back a little nauseous.

For the first time since the industry declared war on made-for-advertising sites, MFA spending went up.

The Association of National Advertisers dropped its Q1 2026 programmatic benchmark in May, and buried in the good-news slideware was a small, ugly reversal. MFA's share of member ad spend went from 0.6 percent in Q4 2025 to 1.1 percent in Q1 2026. The ANA, to its credit, called it what it was: the first meaningful increase since it started counting.

Yes, 1.1 percent is a small slice. Save the "well actually" for someone who cares. Because the trajectory is the story. This is the number that was supposed to only ever go down. The entire post-2023 morality play, the blocklists, the audits, the SSP mea culpas, the LinkedIn thought leadership about "cleaning up the supply chain," all of it was built on the premise that we had turned the corner.

We did not turn the corner. We drove around the block and ended up back at the same dumpster.

And it gets worse if you read past the headline. The advertisers still buying MFA are, shockingly, the ones whose campaigns perform the worst. The bottom cohort spent 2.1 percent of budget on garbage. The top cohort spent 0.9 percent. Which means MFA is not some random tax on everyone. It is a self-inflicted wound, concentrated among the buyers least equipped to notice they're bleeding.

Cast your mind back to that original 2023 ANA report, the one that lit the fire. It found MFA making up 21 percent of impressions and 15 percent of spend, roughly 13 billion dollars a year set on fire annually. The industry gasped. Panels were convened. Vendors were minted. Chris Kane at Jounce and Rocky Moss at DeepSee became the people you name-drop to prove you take quality seriously.

And it worked, for a while. Spend share fell. Fifteen percent, then four, then under one. Everyone took a bow.

But nobody fixed the actual thing. The actual thing is that money follows opportunity, and this ecosystem is structurally, gleefully, allergic to policing itself unless a client with a nine-figure budget makes it. The plumbing that let MFA thrive was never ripped out. It was just told to behave. And plumbing does not have values.

So of course it came back. Of course it did. You cannot build an entire industry on the principle that a fraudulent impression and a real one pay exactly the same, and then act surprised when someone figures out how to manufacture the fraudulent one for a tenth of a cent using a chatbot.

Here is the ANA's own Julie Weitzner saying the quiet part with a straight face: money follows opportunity, and if even a small percentage of advertisers are willing to run ads on these sites, it is worth it to spend next to nothing spinning up domains that might catch a few ad dollars.

Read that again. It is now cheaper to generate a fake website than it is to make a decent one, and the ad ecosystem will pay for both at the same auction. That is not a bug someone forgot to patch. That is the machine functioning exactly as designed.

Remember, ChatGPT opened its doors to the public on November 30, 2022. (Not February 2023, no matter what you read in the trade press, and we will get to the trade press.) In under three years, we went from "neat little chatbot" to "the majority of new web pages are at least partly synthetic." The content explosion did not sneak up on anyone. We watched it happen in real time, congratulated ourselves on the panels, and left the doors wide open.

You want to picture a villain. A teenager in a basement somewhere, a shady arbitrage shop, the usual cast of programmatic gremlins we love to blame because blaming them costs us nothing.

Cute. Wrong.

The slop is not mostly coming from the worst of the worst. A lot of it is coming from the companies whose logos are on the sponsorship banner at the very conference where they'll tell you how much they care about brand safety. It is coming from platforms you already trust, already buy, already put on the media plan without a second thought.

It is coming from Pinterest, a company so overrun with generated garbage that its own CEO went on an earnings call and admitted, out loud, that the platform cannot reliably tell what's real. Think about that, and then think about why you should never buy media on Pinterest.

It is coming from YouTube, where a substantial chunk of what gets recommended to a brand-new user is straight-up slop, monetized through the same AdSense pipes you fund. It is coming from Meta, from X, from the recommendation engines that are supposed to be the "clean," walled, trusted alternative to the open web sewer.

The call is coming from inside the house. And in Part Two, we're going to name every room.