The machine on trial
A federal court in Oakland is deciding whether the attention economy is a public health hazard. Publishers have already learned what happens when Meta changes its mind about them. Everyone else is about to find out.
By Chris Marine, Founder & CEO, Campfire
On August 18, 2026, a courtroom in Oakland opened a trial with a number attached to it so big it stopped working as a number and started working as a headline. $1.4 trillion. That's the ceiling Meta itself has floated for what it could owe if the states suing the company get everything they're asking for in MDL-3047, the federal case now in front of Judge Yvonne Gonzalez Rogers. California, Colorado, Kentucky, and New Jersey are trying the case first, standing in for a coalition of 29 states, in front of an eight person advisory jury that's expected to spend six to eight weeks hearing evidence before a decision lands sometime around October, according to Fortune and NPR. Another 3,137 cases, filed by school districts and individual families, are waiting behind this one.
Nobody serious thinks Meta actually pays $1.4 trillion. A Cornell law professor put it plainly to reporters. He said an award anywhere near that size "would put Meta into bankruptcy, wipe out its owners, and effectively result in the states owning Meta," which just isn't how these things tend to go. Earlier this year, a New Mexico jury found Meta liable on similar claims and landed on $942 million. That's a real number, a painful number, and roughly a thousandth of the one making headlines. The trillion dollar figure is doing rhetorical work here, not predicting an outcome.
But honestly, the number was never the interesting part of this trial. The interesting part is what the states are actually asking Meta to do to the product. According to court filings summarized by Gizmodo, they want Meta to kill infinite scroll. Put real time limits on users under 18. Delete the AI systems trained on children's data. Redesign the core mechanics of Instagram and Facebook so they stop being, in the state's words, engineered to be habit forming. California's deputy attorney general, Megan O'Neill, opened the trial with a three word summary of the alleged business model, "hook, hold, harvest." Her closing line on Meta's internal decision making was blunter still. "Time and again, profits won." Internal documents entered into evidence include a 2016 email describing "teen time spent" as a company goal, and an internal study titled, without much subtlety, "Long Term Retention. The Young Ones Are The Best Ones."
If any meaningful share of those design remedies survives appeal, and legal observers are already comparing the likely outcome to the tobacco Master Settlement Agreement, which traded one ruinous payout for decades of structural restrictions on how cigarettes could be marketed, then the machine that makes Meta valuable to advertisers doesn't just get fined. It gets rebuilt. And a version of Facebook and Instagram that isn't optimized for maximum engagement is a very different advertising product than the one media plans have been built around for the past fifteen years.
Publishers already ran this experiment
Here's the part that should concern anyone whose media strategy still leans on Meta as a distribution partner, whatever happens in Oakland. The platform has already shown the industry exactly what it does when its own priorities shift, and it didn't need a courtroom to do it.
Chartbeat and Similarweb data tracking nearly 800 news and media sites found that Facebook referral traffic to publishers fell 50 percent in a single year, between March 2023 and March 2024, and 58 percent over six years, from 1.3 billion monthly visits down to 561 million, as reported by Press Gazette. Facebook's share of all external referral traffic to those sites dropped from 30 percent in 2018 to 7 percent in 2024. Smaller publishers, the ones without a Times or a Post behind them, fared worst. Sites under 10,000 daily page views are now receiving roughly 2 percent of the Facebook traffic they got in 2018. One UK publisher, Reach, tied a 15 percent year over year drop in digital revenue partly to Facebook's retreat from news. Another industry survey found off platform ad revenue falling 47 percent in a single quarter, with researchers pointing directly at Facebook's decision to deprioritize news content in the feed.
None of that required a lawsuit. Meta simply decided news wasn't worth the trouble, and publishers who'd spent a decade building audience development strategies on top of Meta's feed absorbed the loss. A trial that could force the company to strip out the very features that made the feed sticky in the first place isn't a new threat to that relationship. It's a continuation of one, with the added twist that this time the redesign wouldn't be optional and wouldn't stop at news.
The buzzword that got the diagnosis right and the treatment wrong
There's a term for the discipline that was supposed to protect advertisers from exactly this kind of platform dependency, and it's quietly become a punchline. Supply path optimization. SPO showed up around 2017, in the aftermath of header bidding, when publishers started running the same ad impression through a dozen competing exchanges at once and nobody buying media could tell anymore which path from advertiser to reader was efficient, honest, or even real. The original idea was sound. Figure out which routes to which publishers are genuinely worth taking, and stop paying five different intermediaries to shuffle the same impression between them, as AdExchanger's history of the term explains.
Then the industry did what the industry does with a good idea. SPO got reduced to a fee spreadsheet. Buyers used it to squeeze out the most expensive resellers and called it a strategy, while the real question, whether the inventory on the other end of that "optimized" path was actually reaching a real person in a context worth showing up in, got quietly dropped from the conversation. The Association of National Advertisers has been measuring the results of that for years. Its most recent transparency benchmark found $26.8 billion in programmatic media value lost annually to inefficiency, and a "TrueCPM" efficiency score that actually declined from 37.8 percent to 36.5 percent year over year, meaning the gap between what advertisers pay and what they actually get widened even as the industry got better at avoiding the most obvious junk. Made for advertising spend fell industry wide, but the top quartile of marketers were still routing up to 28.7 percent of their programmatic budget to it, according to the ANA's own numbers. Supply path optimization didn't fail because the idea was wrong. It failed because everyone got obsessed with the dashboard and forgot to ask what the dashboard was supposed to be measuring in the first place, whether the audience on the other end was real, engaged, and worth reaching.
The people closest to the buy side are starting to say the quiet part out loud. One programmatic veteran put the corrected version of SPO pretty simply. There are, at most, a thousand sellers in the entire open web that actually matter to a given brand, probably closer to a hundred, and the job was never to optimize a path through thousands of exchanges. It was to find those hundred and build a real relationship with them. That's not new technology. That's a phone call.
What a trillion dollar trial and a shrinking feed have in common
Zoom out far enough and the Meta trial and the slow bleed of publisher referral traffic are the same story told at two different speeds. Both are about what happens when a business model built entirely on capturing and holding attention runs into a limit, whether that limit is a jury deciding the model is a public health hazard or a platform simply deciding news isn't the attention it wants to hold anymore. Either way, the advertiser who built a media plan on the assumption that the feed would keep working the way it's always worked is the one left exposed. Forecasters are already predicting a rougher road ahead for the exchanges and auctions built on top of that same attention infrastructure, with some industry watchers projecting that agentic buying and margin squeezed retailers could push another 10 to 20 percent of costs out of the programmatic supply chain simply by routing around it entirely.
That's not an argument for abandoning programmatic. It's still, used well, one of the more efficient ways to reach people at scale, and pretending otherwise would be its own kind of lazy thinking. At Campfire, programmatic still makes up just over 25% percent of the media we manage for clients, and we're not walking away from it. But we've watched that share shrink over the past two years, not because programmatic stopped working, but because working directly with publishers, and building real presence with audiences across digital and social rather than renting a few seconds of their attention through an auction, has started outperforming the easier path more often than it used to.
The bulletproof version of a media plan
If the past few years have taught anyone paying attention to the operations side of media anything, it's that the platform you built your strategy around doesn't owe you continuity. Facebook didn't warn publishers before it deprioritized news. A jury in Oakland isn't going to consult your media plan before deciding whether infinite scroll should exist. The only part of a media strategy that survives a redesign, a lawsuit, or a platform's change of heart is the part that was never fully dependent on any single platform to begin with. A direct relationship with the publishers and audiences who were never going anywhere, negotiated on terms you understand, measured against outcomes that matter to your business rather than the metrics a dashboard happened to make easy to report.
That's not a hedge against a $1.4 trillion trial. It's just what a media investment looks like when someone treats it like one. The trial in Oakland will resolve in the next couple of months. The audience you actually need doesn't have to wait for the verdict to find you.
A note on how this was made. This article was researched and drafted with the support of AI, then reviewed and edited by our team before publishing.
