Why Attention Is a Bubble That Won't Burst

On the Italian edition of Tim Hwang's Subprime Attention Crisis (published as La grande bolla dell'attenzione, minimum fax, 2026)

Last take before a short summer pause.

When I was asked to review Tim Hwang’s book, I felt a brief jolt — the way you do when you pick up an essay that smells faintly of heresy and forbidden truth, but also of real insight into an advertising industry going through one of the most seismic transformations it has ever seen.

The book came out in 2020, and today it reads like it was written in the Paleolithic. That’s the first impression it leaves — not because it’s old in itself, but because the pace of change in the media industry, pre- and post-AI, is closer to pre- and post-wheel, or pre- and post-electricity.

The book draws an apocalyptic parallel between the subprime mortgage crisis and the ad tech “bubble,” but with a difference I consider decisive: in the mortgage crisis, default is an event — it arrives on a specific day, when the debtor stops paying. In digital advertising, the default is already here; it’s simply never declared, because the system never measures the real attention people give to the advertising of the new millennium — it only measures their exposure. The day we actually started measuring it, the default wouldn’t surface as a single event. We’d find it spread across every single impression, as a chronic gap between what gets billed and what actually gets perceived.

Five years after the original release of Subprime Attention Crisis, programmatic now handles 89.3% of global display ad spend — over $380 billion, on its way to $800 billion by 2028. The opacity the book describes hasn’t changed; it has simply moved from human intermediation to exchange and bidding platforms, now governed by metrics that measure not attention but hyper-optimized exposure. A large share of brand money isn’t funding a real outcome in attention, sales, or brand-building — it’s funding a giant closed system.

What the book gets right

Credit where due: on the plumbing of the system, Hwang is precise, and the numbers he cites — many updated in the Italian preface itself — hold up well against the present. The ISBA/PwC study of the UK programmatic supply chain, which in 2020 found only 12% of spend fully traceable and 15% vanishing into an “unknown delta” never explained. The J.P. Morgan Chase case, which cut the number of sites it advertised on from 400,000 to 5,000 — a 99% reduction — without campaign results changing by a hair. Facebook’s video metrics, which reportedly overstated average view times by 60 to 80%, by dividing total watch time across a smaller audience than the real one. The same platform sells the inventory, defines the metric, and certifies the result. You don’t even need the bubble theory to see something’s off — you just need to notice who’s grading their own homework.

Then there’s the single data point that says more than any theoretical framework: in 1994, the first online banner ads posted a 44% click-through rate. By 2018, a comparable display ad had fallen to 0.46% — roughly one person in two hundred, and nearly half of mobile clicks were accidental anyway. Thirty years of technical optimization, ever-finer targeting, ever more sophisticated auctions, to arrive at a behavioral return that’s one-hundredth of where it started. If that isn’t a symptom of an industry that stopped producing real attention and settled for producing occasions of exposure, I don’t know what would be.

What the book doesn’t see

And here’s where the problem starts — with a caveat on timing that’s only fair. The original text is from 2020: it couldn’t have cited Netflix or Disney+ running ads, which only arrived in November and December of 2022, nor a creator economy that, as a recognized investment category, emerged around the same time the book was being written. On that, the silence is understandable. Less understandable is that the 2026 Italian edition — with Marco Carnevale’s preface, which extensively updates the numbers on programmatic, opacity, and market share right up to today — doesn’t also widen the conceptual frame. Connected TV already existed as an advertising category in 2020, smaller and more fragmented, sure, but eMarketer was already tracking it with its own figures — and yet six years and a preface later, it still gets a single mention, in a footnote. And then there’s “creativity,” a word with no expiration date: it doesn’t need to be current to be relevant, and yet it never appears once in a hundred and seventy pages, not in the original text, not in the update. That last absence can’t be blamed on the calendar.

It’s an absence that says a great deal, because it isn’t accidental — it’s built into the lens Hwang chooses. If the problem is the financialization of attention — its transformation into an undifferentiated stock, buyable and sellable in bulk — then creativity, context, and the moment an ad appears barely matter at all: only volume and price do. That’s exactly the logic the industry has adopted, and it’s exactly why the book, even while criticizing that logic from the outside, ends up accepting its categories from the inside. It treats attention as though it were purely a measurement problem (which it partly is, as long as the system keeps grading its own homework). No — it was never just that.

Because the real question — the one none of Hwang’s numbers can capture — is a different one: why do we keep seeing, on every screen, out of context, disconnected from the genre of the programme, the time of day, the moment, the same thirty-second spots built for an audience and a medium that no longer exist? Advertising creativity is nowhere near as optimized as the auctions that distribute it. If anything, the more automated the system gets downstream, the more frozen it seems to get upstream. Costs get cut through agency mergers and automation, but the first thing cut is almost always the part of the business considered “unproductive” — creativity — while the teams overseeing data and platforms keep growing. The spreadsheets get more sophisticated, and the content those spreadsheets are supposed to measure gets less care.

Where real attention seems to vote (and where the vote is already rigged)

The point is that the market, without waiting for the “controlled demolition” Hwang proposes as a regulatory fix, is already moving budget elsewhere. Creator economy spend in the US — the only market with genuinely solid data, per the latest IAB report — went from $13.9 billion in 2021 to $37 billion in 2025, and is expected to reach nearly $44 billion in 2026: growing four times faster than the ad market as a whole. Over the same period, global linear TV spend has fallen to $139 billion, its lowest level since 2005, while US digital video spend will pass $80 billion this year — nearly double the market’s average growth rate.

The thing is, connected and digital TV — which promised granularity, transparency, and control compared to old linear TV — may in practice have delivered only the perception of control, while importing linear TV’s same approximations. It’s no accident that almost every statistic describing it as a high-attention environment comes from whoever is selling that inventory.

A survey by All About Cookies of a thousand American adults, conducted in March 2026, finds that only 24% of viewers actually pay attention to streaming ads, while the rest multitask or disengage — and 76% still think there’s too much advertising regardless. Consumer Reports, surveying a representative sample of over two thousand adults, finds that only 3% of subscribers on ad-supported plans don’t find the ads annoying at all: the main complaint is repetition — the same spots keep coming back, despite inventory supposedly being limited. Tellingly, when asked to account for the quality of that attention, Netflix — through its president of advertising, Amy Reinhard — simply stated that viewers pay as much attention to ad breaks as to the content itself, despite having had an independent attention-measurement partner in place for two months by then (Lumen Research, though live in only five European markets), without citing that data to support the claim.

Even the last line of defence — little advertising, scarce inventory, frequency capped by a hard ceiling — is giving way under competitive pressure: Netflix and Prime Video CPMs have both dropped noticeably over the past year, while ad load is increasing, not shrinking. The pattern is the same one playing out in programmatic display, just arriving a few years later: as long as the environment stays scarce, price holds and the attention narrative holds with it; the moment supply widens, the system goes back to behaving the way it always has.

The measurement paradox

There’s one last irony the book brushes against but never develops, and it may be the most useful one for understanding where we really stand. The formats with the most “scientific” measurement standard — viewability down to the pixel, the exact second in which 50% of the ad must stay visible — are also the ones with the lowest behavioral return of all: that 0.46% display CTR, again. The formats that claim to perform better — creator content, connected TV, quality editorial contexts — are still measured with blunt proxies, often produced by parties with an interest in making them look good: engagement rate, completion rate, the old television GRP. Standardizing the measurement hasn’t standardized the value. If anything, the opposite: the more precise the measurement has become, the further it has drifted from what it’s supposed to represent — and where the measurement stays blunt, it’s also easier to tell yourself everything is working.

Hwang calls for an independent authority to audit the industry’s metrics, more transparency, protection for whistleblowers. Reasonable asks — but they assume the problem is mainly one of oversight and rules. It isn’t, or not only. No regulatory body builds a better ad by itself, or the right context to show it in. Something, for what it’s worth, is already moving in that direction without waiting for a regulator — Amazon and Google have just started letting an attention score, still imperfect, weigh on the price of an impression itself, not only in the report that lands once the campaign is over. It’s a signal, not a solution: until building real attention becomes, for whoever is allocating a media budget, the easier choice — not just the more honest one — we’ll keep measuring, with ever more precision, something that keeps being worth less.