On June 12, Google extended its “Limited Ad Serving” policy to Google Search and started a phased rollout that, by the company’s own reported timeline, won’t finish until 2028. That is the actual news behind this week’s front-page complaint that Google is still serving dodgy ads. The complaint is fine as far as it goes. The policy is the thing almost nobody read.

A Throttle, Not a Ban

Start with the mechanism, because it’s the tell. Limited Ad Serving doesn’t disapprove an ad. It caps impressions for an advertiser Google has decided is “unqualified.” The account stays open. The billing stays on. The advertiser just gradually stops being visible — which is exactly the design you reach for when you want to look like you’re enforcing the rules without ever handing back the revenue.

Google will tell you it removed 146 million ads in a single year for misrepresentation. Believe it. That figure is also a rounding error against the volume it approves, and it describes the old regime: catch the fraudulent creative, kill the fraudulent creative, repeat. The new regime doesn’t need to catch anything. It can decide the advertiser is bad and let the algorithm starve him out.

That’s a better business. It’s a much worse system for anyone who isn’t a scammer.

The Complaint Is Now the Verdict

The operative language, as the trade press covering the update reported it: when users have “persistently and disproportionately” reported that an advertiser’s content, products, or behavior don’t meet their expectations, Google “may consider that advertiser unqualified” and limit its impressions.

Read that twice. The input is sentiment. The output is reach. There’s no finding, no specific allegation you can see, no counter-evidence to submit against a particular complaint — the complaints are aggregated into a signal, and the signal is proprietary. StubGroup, a Google Ads agency that publishes its own suspension data and refreshed that data on September 10, reports that enforcement is inconsistent enough for competitors running more aggressive tactics to sail through while ordinary accounts get flagged, and that appeals frequently reach no human being at all.

A media buyer at a mid-size agency, typing into a Slack DM a few minutes after a client’s campaigns went quiet, put it better than any policy document: “They didn’t tell us we were banned. They told us our reach was ‘limited.’ We’ve spent three weeks trying to find out who complained.”

What a Bank Would Have to Disclose

Here is the analogy Google’s lawyers would prefer nobody make. If a bank quietly cut a small business’s credit line because of unverified third-party complaints it refused to itemize, a consumer reporter would be camped outside the branch by Thursday. A statutory framework already exists for precisely that behavior: adverse action, notice, a right to dispute, an actual file you can read. Advertising reach is now a form of working capital for a small business, and it is being rationed by a scoring system with none of those protections — administered by a company that also sells the remedy.

I’m not arguing that Google should be scolded for the sin of screening ads. Screening ads is good. I’m pointing out that “screen the ads” has been quietly swapped for “rank the advertisers,” and the second thing is a power nobody granted and no statute describes. The phased enforcement matters: this is not a pilot that might get walked back. It’s architecture, scheduled for completion in 2028, in the way a zoning change is architecture.

The People Who Should Be Angriest

The scam-ad debate has two camps, and both should be uncomfortable.

The first camp wants platforms left alone and expects self-policing to sort this out. Self-policing just produced a private reputation bureau with impression-level penalties, no disclosure, and an appeals queue staffed by a model. That isn’t the light-touch outcome. It’s the heavy-touch outcome minus the due process, and it should bother anyone who believes private ordering beats public rules — because this is what private ordering looks like when the private party is also the judge, the registrar, and the auctioneer.

The second camp wants Google to “do more” about fraudulent ads. It did more. It did the thing that costs it nothing: it moved from punishing content to punishing customers, and the customers who get squeezed hardest are the ones with thin margins, unglamorous products, and a handful of annoyed reviewers.

The dodgy ads will still be there. That’s the part worth sitting with. The fake crypto exchange that registers a new shell on Tuesday will be back on Thursday, while the gutter-cleaning outfit with a 3.9-star average waits to learn whether its reach has been “limited.” If you want a system that catches fraud, ask for one with findings and appeals. If you want a system that catches the merely unpopular, Google has already shipped it — and it’s on schedule to finish in 2028.

Sources