Hi. Welcome.

Pull up a chair.

If you're reading this on the way to another "AI‑powered future of adtech" breakfast, put the croissant down and sit with me for a second. The croissant isn't going anywhere. Neither is the bullshit you're about to walk into.

There's a particular flavor of self‑deception that defines advertising in 2026, and it tastes exactly like a LinkedIn post from a holding company CEO three weeks after his bonus cleared. It's the flavor of people standing in a burning building telling you they're "excited about the transformational heat opportunities" this presents for their proprietary AI‑powered firefighting platform. Deck available on request. Series C is open. Lead investor is a guy who used to run programmatic at a company that no longer exists, for reasons nobody discusses in polite company.

I've been doing this long enough to remember when "programmatic" was a promise. Efficiency. Transparency. The right ad, right person, right time. A lovely sentence. A lovely decade of lovely sentences. And here we are in 2026, with roughly thirty‑six cents of every programmatic dollar actually reaching the publisher who produced the content the ad is allegedly sitting next to, and the industry still has the nerve to call this a "temporary inefficiency."

If a restaurant told you sixty‑four cents of every dinner dollar went to a guy in the alley who may or may not exist, you wouldn't call that supply chain optimization. You'd call the police. In adtech, we call it Q4 and we give the guy in the alley a panel at Cannes.

Global ad spend is cruising toward $1.3 trillion in 2026. The money isn't going anywhere new. Same toll collectors, faster, now with an AI wrapper so the invoices can be 20% higher and nobody can read them without a PhD and a priest. The story of 2026 isn't expansion. It's extraction wearing a Patagonia vest.

Below: all five hard truths. The diagnosis is free. The prescription, how to actually make money off the wreckage and not end up quoted in someone else's postmortem, lives in Part Two, behind the ADOTAT+ paywall. Because Chase Bank does not accept "exposure" as a form of payment, and neither do I, and neither does my rabbi.

The advertisers worth taking seriously have stopped asking how to optimize digital. They're asking whether digital‑heavy strategies still work at all. That's the question you ask about a foundation, not a paint color. That's the question you ask right before you call a structural engineer, or a lawyer, or both.

Rising CPMs. Shrinking attention. Collapsing trust. Auction pressure intensifies. Inventory expands like a balloon animal at a six‑year‑old's birthday party, and about as load‑bearing. Brands are spending more money to reach fewer people who trust the message less. If you pitched that as a business plan at a board meeting, someone would ask you to leave the room. In adtech, we call it "scaling."

This isn't cyclical. It's structural. The open web is being asphyxiated by AI answer engines that eat publisher traffic before it ever arrives. Walled gardens are absorbing the survivors like a sponge absorbs a small, confused puddle. And the one thing everyone at every conference agrees on is that measurement is broken — which is, frankly, a fascinating thing to unanimously agree on while simultaneously reporting record efficiency numbers to your board on Tuesday.

Measurement is broken. But the dashboards are green. Both cannot be true. Somebody is lying. It's probably not the measurement. It is almost certainly the guy who built the dashboard, and he is almost certainly about to get promoted.

Digital was supposed to be the accountable channel. The grown‑up one. The one where we'd finally know which half of the advertising was wasted. Plot twist: it's both halves. It was always both halves. We just built dashboards pretty enough to hide it and hired people charming enough to present them.

Here's the sentence the IAB would very much prefer I not print, so obviously I'm about to print it in 36‑point bold and send it to their press office with a smiley face:

Lower fraud would directly shrink reported media revenues. That is why fraud persists.

Read it again. Slowly. That's the whole game in one line, and I apologize in advance for the fact that you can't un‑see it. That sentence is going to ruin at least one lunch you have with a vendor this quarter.

Juniper projects global ad fraud will top $100 billion in 2026. MFA sites eat 21% of programmatic impressions, per the ANA's own study, which they commissioned, which they published, and which they then responded to by holding a press conference, nodding solemnly, and changing approximately nothing. In finance, home services, and legal, invalid traffic rates hit 42%. That's almost half your traffic being ghosts. Friendly, click‑happy, conversion‑mimicking ghosts, pouring your budget into the void while a dashboard cheerfully reports growth and your agency lead talks about "momentum."

The industry's response to all this? Invent a new fraud detection vendor every six months. Charge it back to the advertiser as a "brand safety line item." Put someone charming on a panel to talk about "the industry's commitment to quality." Go to dinner. Expense the dinner. Keep the party going until someone turns the lights on, which they never will, because they own the light switch.

Fraud is the ad tax. A baked‑in, cost‑of‑doing‑business skim everyone complains about at dinner and absolutely nobody disrupts during office hours. Because disrupting it means admitting that a meaningful percentage of the inventory your entire P&L is built on does not exist, has never existed, and is being invoiced to you in clean whole numbers with a logo on top.

The honest questions for 2026 aren't "what's my ROAS." They're: How much of my spend reaches real people? How much of my performance reflects actual human attention? And if I found out the real answer, would I still have a job on Wednesday?

(Spoiler: probably not. That's why nobody asks.)

Where Your Programmatic Dollar Actually Goes in 2026 (Or Doesn't)

"The 44‑cent figure is the industry's current best case. The 36‑cent figure is what actually lands with the publisher who wrote the article your ad is next to. Everyone between the dollar and the destination is taking a cut, a vig, or a commission. Most of them are doing it with a LinkedIn post that says 'excited to share.'"

Matt Spiegel said it cleanest on The ADOTAT Show, and I've been stealing it ever since with full credit and zero shame: AI is not the solution to your performance problems. It's an amplifier of everything already broken underneath. Siloed data. Incomplete identity. Attribution held together by string, hope, and a Google Sheet nobody has owned since 2022. AI doesn't fix any of it. It exposes it. Loudly. At scale. With the confidence of a man who has never once been wrong in a meeting.

The vendor demo shows the AI writing 300 creative variants in four seconds. What the demo does not show is the AI taking your broken attribution model and confidently optimizing toward a metric that was never real in the first place. Just faster. With fewer humans in the loop. With fewer fingerprints. With a much nicer PDF report at the end. The metric isn't real, but the invoice is, and the invoice clears.

And then there's the creative layer, which is where things get properly dystopian. Generative tools have made deceptive advertising cheaper than at any point in human history. Any guy with a laptop, a free trial, and vibes can now produce a video ad that looks like a brand campaign, and a lot of those guys are using that capability to push supplements that don't work, crypto that isn't crypto, and scams your mother is one unguarded afternoon away from falling for. This is the next evolution of internet spam, and it's wearing your creative director's haircut.

The downstream consequence is already showing up in consumer sentiment: audiences are starting to distrust anything that looks AI‑made. Including your legitimate brand spot. Including the campaign your agency is quietly proud of. The brands that adopted generative AI fastest for efficiency reasons are about to discover they bought a trust penalty along with the productivity gain, and nobody told them because the invoice didn't have a line item for "brand erosion."

The AI‑washing bubble is real, and 2026 is the year it gets pricked. The honest question isn't whether AI changes advertising. It's whether you can tell the difference between AI that reduces waste and AI that automates waste at lower headcount with better branding. Most people, as of this sentence, cannot. Which is fine. That's why Part Two exists.

(How to actually tell the difference, and the audit to run before you sign another AI vendor contract, is in Part Two. Behind the paywall. I know. I'm sorry. Not that sorry.)

The rule for 2026 is short enough to tattoo, and some of you should:

If a platform cannot show you traffic origin, placement logic, algorithmic decisioning, and data flows, it does not get the budget.

That's the whole rule. It is violated, hourly, by every major holding company on earth, because the alternative is admitting that a lot of the "premium" inventory they've been buying for you can't actually be audited to anyone's satisfaction, least of all a client's, least of all yours.

Transparency in 2026 is the last real competitive edge, which is precisely why so few players offer it. MFA noise. Black‑box placements. Opaque attribution. Walled‑garden dashboards you're expected to accept on faith like you're at a slightly dishonest synagogue. Publishers who refuse to expose page‑level data. DSPs quoting effective CPMs that are mathematically impossible to verify and then looking personally wounded when you ask them to verify it, as if you've insulted their mother instead of questioning their arithmetic.

CTV is the purest case study in the grift. "Premium" has quietly come to mean "more expensive," not "better verified." A $75 CPM on inventory that turns out to have been served into a picture‑in‑picture corner of a free ad‑supported channel nobody was watching is not premium. It's a rounding error with good branding, a nice conference booth, and a VP of Sales named Brad.

The buyers who win in 2026 and 2027 treat clarity as a hard filter, not a nice‑to‑have. The rest keep funding the chaos and then expressing shock, once a quarter, when the chaos behaves exactly like chaos. It's like watching someone marry a known arsonist and then act surprised when the kitchen catches fire.

(The four‑question transparency audit that forces every vendor in your stack to either show the work or lose the budget by Friday is in Part Two. Yes. Still behind the paywall. We have been over this.)

Here's the one nobody wants to say out loud at a conference, mostly because half the conference's sponsors have a vested interest in nobody saying it.

The open web is facing genuine, existential, not‑a‑drill pressure. AI answer engines are absorbing the consumer traffic that publishers built their businesses on. Referral traffic is collapsing. Search queries are being intercepted and answered before they ever reach a publisher's page. Ad spend is migrating, aggressively, into walled gardens, CTV, and commerce media. The survival strategy for serious publishers is a full retreat under fire into subscriptions and first‑party data, because everything else is being hollowed out in real time and the hollowing is being described, cheerfully, by the people doing it as "industry evolution."

And retail media, which has been sold to the industry for three straight years as the savior of post‑cookie advertising, is quietly failing to live up to the pitch for approximately everyone whose name is not Amazon or Walmart. The entire ecosystem of third‑party commerce adtech and hundreds of retailers piled revenue expectations behind the category like chips on a single number at a roulette table. The actual spend has accrued, overwhelmingly, to Amazon, then Walmart, and then to a thin, embarrassed trickle across everyone else. Kroger, Target, Best Buy, every grocery chain that launched a retail media network in the last 36 months with a PowerPoint deck and a dream: "build it and they will come" was not a media strategy in 1989, and it is not a media strategy now. Somebody should have told them. Somebody probably did. Somebody was ignored at the offsite.

Over the next 24 to 36 months, every retail media network that does not make a deliberate strategic choice about what it actually is will have that choice made for it. By advertisers. By consolidation. By quiet irrelevance. Probably all three in roughly that order, and each with its own wave of LinkedIn posts about "new chapters" and "exciting next steps."

The Retail Media Reality Check: Who's Actually Cashing the Checks

"Dozens of retail media networks launched since 2022. Roughly three are meaningfully profitable standalone businesses. Your CFO should be modeling which of the rest are going to be press‑released out of existence between now and 2028. Start with the ones whose general manager has updated their LinkedIn twice this year."

The real 2026 breakthrough isn't coming from another shiny targeting feature or another AI optimization wrapper with a snappy name someone workshopped in Brooklyn. It's coming from the unsexy, unfundable, unexciting work of fixing the stubborn structural problems that still let MFA eat 21% of impressions, still let fraud grow 14% a year, and still keep billions of TV dollars stuck on the sidelines waiting for someone to be honest.

The industry is being stress‑tested from every direction at once. AI pressure. Macro pressure. Trust pressure. The myths that held it together for a decade — omnichannel, clean buy‑side/sell‑side separation, the comforting assumption that programmatic was working roughly as advertised — are all being forced into reassessment at the same time, by the same people who built them, with the same straight faces.

You just read what's broken. Part Two is what to actually do about it on Monday morning, including:

[Subscribe to ADOTAT+ →]

Your CMO can expense it. Your agency absolutely cannot. That asymmetry, honestly, is the whole point.