Massive jump in #GPC adoption in the last week. Here's sites in Well-Known with GPC support resources over the last week.

Note that only 1,446 say they *do* support GPC - the rest are declaring they *don't*. But that's still a 10x increase.

(cc @globalprivctrl @robinberjon) @globalprivctrl @robinberjon The vast majority of the new sites are hosted on the 23.227.32.0/19 subnet, which is allocated to Shopify.

So it looks like Shopify has added a GPC implementation to their platform, rather than this being a broader adoption uptick.

Still, getting Shopify onboard is big!

Of the ~1M ads.txts Well-Known tracks, 99% authorize at least one Google account. Over 97% authorize at least one Google account as DIRECT.

The next closest ad system is Magnite at 24% (19% DIRECT). Looking at just sites in the top 100k (based on Tranco rank), over 97% authorize at least one Google account (93% DIRECT).

Once again, the next closest ad system is Magnite, with 73% (61% DIRECT).

Of the ~1M ads.txts Well-Known tracks, 70% only authorize a single account to sell their programmatic adspace.

Of those 700k+ sites, their single account is from Google over 99% of the time. The next largest ad system is MGID, with just over 1k sites.

well-known.dev/?q=resource%3A… Looking at just sites in the top 100k (based on Tranco rank), far less only authorize one account - 20% of the 19k ads.txts.

But the pattern is similar - it's a Google account 95% of the time.

Let me introduce you to srbin[.]info, a Serbian site that publishes a mixture Russian war propaganda, anti LGBT hate, conspiracies about the "deep state"/George Soros, etc.

With top stories like this, brands wouldn't be advertising on it, right? Right?

🧵 1/n But of course they are!

There are ads plastered all over the site, and while some are bottom of the barrel, there's some pretty big brands too.

Here's the full version of the above image. There are ads for @amazon pharmacy, and a particularly unfortunate ad for @nielsen.

2/n

An aspect of the current sellers.json implementation mess I haven't talked about much yet is that a lot of ad systems go by different domains in different places.

This isn't great when the whole ads.txt/sellers.json/SupplyChain system is based on matching domains.

1/n There's three things that need to line up:

1. A specific domain is used to list accounts from the ad system in ads.txts,
2. That domain hosts a sellers.json which contains those accounts,
3. The ad system's accounts with other ad systems use that as their owner domain.

2/n

Adtech head-scratcher for the day: what on earth did Mintegral just do to its sellers.json?

9,997 changed sellers - only 29 left unchanged. The vast majority were new additions - 8,496 of them. But only 6 have a name or domain.

The rest have no ownership details, despite *not* being marked confidential.

I want to dig a bit more into why removing seller domains is problematic.

Here's 5 accounts that recently had their domains removed. Three are for Russian propaganda sites. Two are for conspiracy/pseudoscience/disinfo/etc. sites.

1/6

https://twitter.com/braedon/status/1540168004830232576

The primary way to track which adtech companies work with "problematic" sites like this is via the seller domain field.

There are sometimes other ways to work it out, but they're a lot more manual, and the links are often more deniable. (Perhaps the point of the removals?)

2/6

Well-Known can now validate SupplyChain objects against ads.txt, app-ads.txt, and sellers.json data.

Lets go through an example, using some SupplyChain objects taken from header bidding requests.

well-known.dev/resources/ads_… Supply Chains can be in JSON form (each formatted on a single line) or the serialised form used in URL params.

The optional publisher domain field enables validation against ads.txt.

Note that there's an app-ads.txt version of the validator as well: well-known.dev/resources/app_…

It's crazy that OpenWeb still has the wrong domain listed for these accounts, almost 4 months after @nandoodles and I first flagged it to them.

They supposedly have a highly vetted set of premium publishers, but they don't seem to have checked they even have the right site?

https://twitter.com/nandoodles/status/1463393702047408138

Here's my thread from August on these accounts, tagging OpenWeb.

Note that at that point, they had 7 youth.conservatives[.]com accounts.

There's now 9: well-known.dev/resources/ads_…

https://twitter.com/braedon/status/1423519068452052995

Here's something interesting. The Publisher Desk, an SSP that's known for working with a lot of disinformation sites, seems to be cleaning house.

Here's the sellers on the Iffy Index that have been removed from their sellers.json.

https://twitter.com/braedon/status/1419659467528933385

Included in there is the account for Steve Bannon's populist[.]press. This account was how he used to access OpenWeb (though OpenWeb had dropped The Publisher Desk earlier).

https://twitter.com/nandoodles/status/1416891528774144002

This whole conversation between Justin and Robin is incredibly frustrating to read, and reinforces my lack of trust in Google/Chrome around #privacy issues, particularly in relation to their #PrivacySandbox proposals.

https://twitter.com/robinberjon/status/1373691498240339975

A little bit of context first - this isn't just some argument between random dudes on the internet.

Justin was "Engineering Director on Chrome Security & Privacy" for the last 11+ years, retiring just under a month ago.