When Guy Parker sat down to write the foreword to his organisation's annual report at the end of 2025, he reached for a pair of numbers. A decade earlier, they would have described a different institution altogether. In 2012, the Advertising Standards Authority, the body that polices what British advertising is allowed to claim, spent, in his words, "5% of our regulatory resource on proactive work and 67% on reactive complaints casework."7 It waited for the public to notice a dodgy advert and write in. By the end of 2025, he wrote, "that had shifted to 45% on proactive work and 38% on reactive complaints casework."7 The watchdog had spent thirteen years quietly turning itself inside out.
The instrument of that reversal was a piece of software the ASA built in-house and named, without much ceremony, Active Ad Monitoring. It is a system that reads adverts at a rate no room full of people could match. In 2025 it "processed nearly 60 million ads,"7 surveying, sorting, and flagging across more than thirty regulatory projects. A regulator whose founding tool was the public's complaint had, in effect, grown a second sense. It could now go looking, on its own, at a scale a person cannot imagine.
A machine now finds the adverts. What has it been finding that no person ever reported?
The regulator no longer waits for complaints
The answer arrived, in the closing weeks of 2025, as three lines of clothing advertising.
For most of the ASA's six decades, a complaint was the only way an advert got caught. The system's own briefing describes the old method plainly: "in the past, we would have relied on limited, labour-intensive manual searches and complaints from the public to stay on top of any non-compliant ads."6 A complaint needs a member of the public to see an advert, recognise something wrong with it, and take the trouble to write in. The adverts that got caught were, overwhelmingly, the ones a great many people saw and could point to: the poster on the Underground, the television spot, the full-page spread. Active Ad Monitoring broke that dependence. It does not need anyone to notice.
Every week the system goes to work. In the ASA's own words, it is built to "capture and sort through large volumes of ads, on a scale humans simply can't,"6 and it hands the ASA's own experts an organised list of the ones it thinks are worth a human's attention. A person then reviews each flagged advert and decides whether to open a case. Through that monitoring, three paid-for Google search adverts were flagged for a human to review, all selling clothing, all making the same shape of green claim: Uniqlo, adidas, and Calvin Klein.
By the summer of 2026 all three had been ruled against and upheld. And on the public file of each ruling, in the box that records how the case began, sits the same small figure: "Number of complaints: 1."1,2,3
That "1" is not a shopper. Each ruling carries the same sentence explaining where the case came from: "The ads were identified for investigation following intelligence gathering by our Active Ad Monitoring system which uses AI to proactively survey ads in specific sectors."1,2,3 The single complaint, in every case, was the regulator's own. It was the machine's referral, counted as a complaint because the system has no other box to put it in. Not one member of the public sent any of these three adverts in.
The ASA's Chair, Baroness Morgan of Cotes, drew the line herself in her half of the same annual report, naming "both the advertising referred to us, as well as the advertisements we proactively monitored."7 Two categories, side by side: the adverts the public hands to the regulator, and the adverts the regulator now has to go and find. These three fell entirely into the second category. Nobody in the public handed them over. Something about them made them almost impossible to hand over.
One of the adverts shows what that something is.
What a paid-search ad actually is
Type a query into Google and, above the ordinary results, sit the paid ones: the adverts an advertiser has bid to place there. One of Uniqlo's, for its fleece range, made its whole environmental case in two words: "Recycled Materials."1
Eighteen characters. That is not a stylistic choice; it is close to the ceiling. Google's own documentation for these adverts states that "the headline fields for responsive search ads support up to 30 characters."4 A paid-search headline is a slot roughly the width of a short shop sign. The machinery behind it is stranger still. The advert you see is assembled for you, in the instant you search, out of parts the advertiser supplied. Google's help pages describe an auction as "the process that happens with each Google search to decide which ads will appear for that specific search."4 The ranking that decides whether Uniqlo's line appears above or below a rival's "is calculated every time a user does a search and is recalculated for different positions on the search results page,"4 based on "several factors, including your competition, the context of the person's search."4
The whole story fits in a sentence. The advert is built for one person's one search, ranked fresh against whatever else is bidding, and gone the moment the page refreshes.
The phrase "assembled for you" is literal. Two people searching the same three words in the same minute need not see the same advert in the same place. Which line surfaces, and where it ranks, turns on "the context of the person's search" and the state of the auction "at that moment."4 A results page is not a public notice board that everyone walks past and reads the same way. It is a private view, built for one search and taken down when that search ends.
Two entirely separate things are true of "Recycled Materials," those eighteen characters on a results page, at the same moment. You cannot tell how much of the coat is recycled: the whole garment, or a fifth of the shell. And no second shopper will ever see the exact line you saw, because it was assembled for your query and dissolved when you clicked away. One surface, and two different ways it leaves you in the dark: you can't size the claim, and you can't witness it twice.
Those two failures do not have the same cause. The size problem comes from the thirty-character box. The witness problem comes from the fact that the advert is personalised and momentary. A billboard is just as cramped for space and yet anyone can photograph it. A long personalised email has all the room in the world and still vanishes into one inbox. Both properties sit on the same surface at once. It is a surface built, by design, to be a fleeting private match between one searcher and one advertiser, not a durable public notice. That design is exactly what makes it a poor fit for a complaints system that runs on members of the public noticing the same advert and writing in about it.
Reporting a paid-search advert is not impossible in the strict sense. A determined shopper could screenshot the line, note the advertiser and the words she had typed, and file a complaint. The ASA accepts complaints about paid search like any other medium. But "possible for a determined person" and "reachable by the complaint system" are not the same thing. That system works when a great many people see the same durable advert and a handful are moved to act on it. It goes quiet when the advert is a private, momentary match that no second person can be shown. The file on all three rulings records "Number of complaints: 1," and that one was the regulator's own machine. The surface is not un-reportable in law. It is un-reportable in practice. For a body built, for sixty years, on the public's eyes, the two amount to the same blind spot.
The first problem is still open. If you cannot tell how much of the coat is recycled, what does the word "Recycled" behind it actually guarantee?
How little "recycled" is allowed to mean
Less than most people would guess.
The most widely used scheme behind a "recycled" garment is the Global Recycled Standard. According to SCS Global Services, one of the bodies accredited to certify against it, the standard covers "the trading and distribution of all products made with a minimum of 20% recycled material."9 Twenty percent is the floor at which a product can enter the system at all. The threshold for putting the consumer-facing GRS logo on a finished item is higher ("the recycled content must be at or above 50%"9). But the two numbers are close enough, and buried deep enough, that they are routinely confused. Textile Exchange, which owns the standard, says as much in its own list of frequently asked questions: "in the GRS, it mentions that the minimum content percentage is 20%, but in the Standards Claims Policy, it says the minimum is 50%."9
Twenty percent is not a trick. A coat that is one-fifth recycled polyester genuinely contains recycled polyester, and Uniqlo's fibre was real. It was post-consumer recycled material, third-party certified. The failure here is one of shape, not of fabrication. The word "Recycled," entirely true, is compatible with a garment that is mostly not recycled at all. And that is precisely where the reader's mind and the certificate part company. A shopper reading a bare "Recycled" does not reach for the floor. They reach for the whole coat.
The ASA reached the same conclusion, and put it in law. In the Uniqlo ruling it held that "as the ad did not include information explaining the basis of the claim 'Recycled Materials', we considered consumers would interpret the claim as absolute," meaning "that all of the fabrics used in the fleece coats and jackets referred to in the ad were made entirely from recycled materials."1 An unqualified green adjective, standing alone, is read as covering everything. The substantiation then failed on exactly that gap: the ASA "acknowledged that each of the products listed on the landing page contained some recycled materials," but found that "the evidence did not demonstrate that each of the products was entirely made from recycled fabrics."1 The clothes contained some. The word promised all.
The code puts the burden of that promise squarely on the advertiser. Before running a claim, a marketer "must hold documentary evidence to prove claims that consumers are likely to regard as objective."5 For a green claim the requirement is sharpened twice over: "the basis of environmental claims must be clear," and "absolute claims must be supported by a high level of substantiation."5 A lone "Recycled," read as covering the whole garment, is an absolute environmental claim standing on a surface that carries nothing beside it to make its basis clear. It was set up to fail that test the moment the number was left off — not because the underlying fibre was fake, but because the word, alone, said more than the evidence behind it could hold up.
The adidas ruling turned on the same hinge. Because that advert referred to a "Range," the ASA held the line "Recycled Running Shoes" was "likely to be understood as meaning that all shoes in their recycled running shoe range were made from 100% recycled materials."2 The company contested that reading. Its answer was that a shopper would read the phrase only as meaning the shoes "contained materials derived from recycled sources," and it noted the brand "did not … operate a standalone recycled running shoe range."2 A reasonable reading, honestly offered. The regulator's point was that the bare adjective does not carry it: left alone, "Recycled" is heard as absolute, and an absolute claim has to be true of everything.
So a truthful version of any of these claims needed a number, or a scope, to pull it back down to earth. The advertiser offered one defence, and it turns on the box.
The claim that would have fit the box
Calvin Klein made the argument out loud, and it is a sympathetic one. The company "acknowledged that the claim, as presented in the constrained format of a paid-for Google search ad, had not included the level of detail that would most precisely have communicated its scope."3 The advert was in a tight little box, and the box did not have room for the fine print. Anyone who has tried to say something careful in a headline will feel the force of it. The advertising code itself seems to leave the door open, requiring that a marketing communication "must not omit material information" while adding that "regard will be had to any limitations (e.g. as to time or space) resulting from the means of communication used and steps taken by the marketer to overcome these by providing the information by other means."5 Time or space. The paid-search box is nothing if not short of space.
This investigation continues below.
Want the next one in your inbox?
The ASA closed that door in a single sentence. It "considered that the ad was not limited by time or space to such an extent that the information could not have been provided."3 The format was tight, the regulator agreed. But it was not so tight as to make the truth impossible to tell. And the code's own escape clause names the way out: the missing detail can be supplied "by other means," which for a search advert means the landing page one click beyond the headline. The box was never the reason the claim overreached.
The box actually has plenty of room. "Recycled Materials" runs to eighteen characters and leaves a dozen to spare inside the thirty-character limit. A figure fits easily: a string like "42% recycled polyester" is twenty-two characters, comfortably inside the ceiling. What genuinely will not fit is a fully qualified scope: something like "42% recycled polyester, shell only" runs to thirty-four, over the line. The most precise possible caveat is squeezed out by the format. But a number is not, and neither is a claim that simply declines to imply "entirely." Character count is not the barrier.
That "42%" is an illustration, not a fact. These adverts did not sell one coat with one recycled percentage. They sold ranges. Calvin Klein's certified recycled content spanned anywhere from twenty percent to a hundred across its products. Across a range like that there is no single true number to print: "42%" describes nothing real, and no advertiser here omitted it. That is exactly why the case does not rest on "the honest figure would have fit." It rests on what the regulator actually said: the advert did not have to make the absolute claim in the first place. It could have scoped the word, pointed to the detail on the landing page, or simply not implied that everything was recycled. The remedy the ASA pointed to was never "print the percentage." It was "don't say more than you can stand behind." The thirty-character box selects for the short bare adjective. It does not force it, and it does not excuse it.
All three of these adverts were, in the end, caught. What did the catching actually fix?
What the machine reaches, and what it can't
The genuinely good news is real, and it fences off the cheap version of this story. Where the ASA has swept a sector and ruled, the adverts largely fall into line afterward. The system's own reporting gives the clearest figure available: after its 2023 airline rulings on misleading "sustainable" and "eco-friendly" claims, follow-up monitoring of around 140,000 adverts turned up only five non-compliant examples, and those five came from a single advertiser.6 Five in a hundred and forty thousand. That is not a picture of a market drowning in lies. Where the machine looks, and after it has ruled, compliance runs very high. And this is not a story that "greenwashing is everywhere," because on the one large piece of evidence available, in a monitored sector, it plainly is not.
The whole system, remarkably, runs on the advertisers' own money. The ASA takes "no direct Government funding,"8 and is instead paid for by "a 0.1% levy on their display advertising spend,"8 a levy that media agencies have also "applied … to advertisers' paid search advertising."8 The surface no shopper can police is, in a small way, helping to fund the machine that now polices it.
That good news has a gap in the fence, and no compliance rate can close it. The reason those airline adverts now behave is that the machine looked. It is the machine that looks, not the shopper standing at the results page. She still cannot size the claim in front of her, and still cannot witness it a second time to check or report it. A green word with no number does not tell you the advertiser is lying. It tells you something narrower and more permanent: that you cannot tell how much, and that the format is the reason the number is missing. If a figure exists, it lives on the landing page one click on, where the regulator itself says the basis belongs, not in the headline, which was built to be too short and too fleeting to hold it.
The three rulings quietly record that collision. One surface was engineered to be seen once, by one person, and then to disappear. Another was engineered to see everything, sixty million adverts a year. For most of the ASA's history only the first existed, and a claim served on a disappearing surface was, in practice, a claim no second shopper could report. The machine is what finally witnessed the three that no one else could. It caught them one advert at a time, in a specific sector, in a particular week. And the box goes on favouring the short bare adjective as surely as it did before, because that is simply what the format rewards.
The machine is already choosing its next sector. Having swept fashion, the system turned to air travel: in July 2026 the ASA upheld rulings against Eurowings and Qatar Airways (the Eurowings advert a paid-for Google search ad), both "identified for investigation" by Active Ad Monitoring as part of "a wider piece of work" in that sector.10 The pattern is the one already seen in the clothing cases: a green claim on a paid-search surface, found by the machine rather than sent in by a shopper. The shape of the problem predicts exactly that. A claim gets checked when the sweep reaches its sector, and not before. The shopper cannot summon the sweep, and cannot reliably even show anyone the advert that unsettled her. Whether the green word in front of you is ever examined depends on where the machine is pointed next, not on anything you are able to do about it.
The reading would change under one condition. If the machine's reach grew until it swept every sector, every week, the shopper's inability to police the surface would stop mattering: the regulator would have become her eyes, and the structural blind spot would close. It has not yet. Coverage still depends entirely on where the AI is pointed next, the adverts still arrive one personalised query at a time in numbers no published sweep has counted, and the surface remains one where the old mechanism (a member of the public noticing and writing in) cannot reach.
Guy Parker's watchdog grew a second sense. It is a real one, and it caught three adverts that would otherwise have vanished unremarked. But reaching the surface is not the same as reaching the format, and neither is the same as handing the shopper back her own ability to check. The next time a green adjective turns up alone in a search advert, with no figure beside it, that absence is the thing to notice. It is not proof that anyone lied. It is the sign that the one number that would let you size the claim is the number the box was built to leave out.
The Magic Wand