@This_Liss @BishPlsOk yes but i don't know that publishers _want_ AI visiting their sites. There's no utility for them right now, and no paid utility on the horizon.
English
Gareth Hates AdTech
1.7K posts
Gareth Hates AdTech
@HatesAdtech
Just another cog in the ad tech wheel.
New York, NY Katılım Mayıs 2010
223 Takip Edilen614 Takipçiler
@HatesAdtech @BishPlsOk It's just like ads for humans. As a publisher, if your ad load is too invasive, irrelevant, and annoying, humans stop visiting your site. Same for AIs. You'd lose domain authority inside AI search APIs like Exa, Parallel, etc.
English
Okay, let's talk seriously about themobian.ai/aboutagentads.
We beat them to milestone of "first ad served to agents" (stevenliss.substack.com/p/how-to-win-f…) and it's both a fascinating pivot and ~inevitable direction for the market, as I see it.
There are some hurdles to scaling, though: (🧵)
English
tbf, if i'm a publisher, my response to this is kind "at worst for whom?"
Maybe the best defensive strategy for publishers is some kind of captcha system that makes it very difficult for AIs to discern the value in the act of crawling, but easy for humans to read it. Holy fuck.
English
Yeah, Claude does *not* like being advertised to. The other models are more chill. Grok is kind of down for it.
The difference between cloaking for SEO and ads for agents is the agent already filters for only relevant content. If you try to inject unrelated sponsored info it rejects it as irrelevant at best, or doubts all the organic content at worst.
AI retrieval is an always-on, soft ad block.
English
@BishPlsOk @aripap @Conductor the value of its attention is not in question -- it's CERTAINLY valuable -- but the cat and mouse game of modifying the behavior of algorithms by manipulating training data leads to worse algorithm performance. Is it an inevitability? Yes. Is it gross? Also yes.
English
@HatesAdtech @aripap @Conductor x.com/BishPlsOk/stat…
The real measure of peer-sapience and moral consideration of an entity is whether the market prices its attention as economically valuable! 😊
More seriously: ~important to delineate between scraping for training and web search / non-learning retrieval.
Michael Bishop@BishPlsOk
Kind of guy who advocates for the moral patienthood (if not equivalence) of AI to open up up net-new surfaces for advertising -- nonhuman but valid traffic deserves to have its "attention" monetized too!
English
@aripap @BishPlsOk @Conductor Seo lead to traffic, getting used by an ai crawler right now doesn’t lead to much (yet). But prompt injecting ai while it’s trying to learn is dystopian, better to just make them pay than to make your content less valuable to them.
English
@HatesAdtech @BishPlsOk I remember @Conductor launching this for "organic" SEO circa 2010.
My biggest concern is that you are potentially hurting the SEO/AIO for the organic content since your page is now "about" something else.
English
In what universe does the misaligned version not utterly dominate the landscape? It’s literally like writing positive words and high value keywords in invisible text in your page.
That and hiding ad infrastructure is basically impossible.
Also, I would think that of all things ai would get good at recognizing when this was happening to them, but I might be giving those nerds too much credit.
English
The aligned version of ads for agents is sponsored data that surfaces info and offers from the deep web that the AI would not have found otherwise. Or info that would’ve been too long to include in the human-readable publisher page.
The misaligned version (that will also happen) is Internet scale prompt injection attacks.
Google/Anthropic/OpenAI’s main reason to allow this is to avoid getting blocked entirely by publishers. Pubs need an economic model to replace lost search traffic, and pay-per-scrape and data licensing isn’t cutting it.
English
@This_Liss @heyitsalexP (In the context of advertising that is)
English
@This_Liss @heyitsalexP There’s a philosophical conversation about art to be had here. While we often ascribe artistic ability to “taste,” that is probably just our monkey brains dealing with a confluence of luck and a lack of deep introspection and analysis.
English
We need to replace "taste" in the AI marketing discourse with "efficacy"
efficacy =
1. the ability to pick winners
2. view the totality of creative/content output through an editor's lens, where you are working in service of your target customer vs some aesthetic ideal
Jess @ FireTeam 🔥 An agency on fire...@HireFireTeam
I'm getting a little tired of the "humans have taste and AI doesn't" argument. 1. Being a human does not ship with 'taste'. I can count on one hand the number of DTC marketers that actually have taste of value. It's a real skill, not a default human quality. 2. You know what I DON'T see present in most of the top ads in ad accounts? Taste. The amount of mashup-AI-voice-over-strategy-slop that I see take the top slot in an ad account is wild. Taste driven marketers vs mindless scrolling consumers is a mis match on the card. I DO think there are plenty of human qualities that supersede AI abilities, but I don't think taste is the right attribute (or right word) for this space. But like 95% of DTC takes, this is probably just semantics.
English
@mikeroplastics @Drone_Wars_ Yeah I saw another discussion about this in a robotics thread wrt one of the 4 limbed/wheeled dog robots. They really went whole hog with making this thing look demonic, like fuckit what if we made it actively scary instead of just uncanny.
English
@Drone_Wars_ @HatesAdtech reminds me of our discussion a while back about how human size/shape robots would be less efficient in most cases than giant robots with many arms and legs, but too horrifying for people to accept right away.
Curious to see how this plays out
English
@ReReadingWolfe If everything is performative nothing is performative.
English
@TimVHook @101Programmatic I could get real philosophical on this one, but I’ll save it for beers. I certainly think that ai’s “alien-ness” will be short lived. Chatty is already fundamentally woven into my day to day.
English
@HatesAdtech @101Programmatic It’s reminding me of when photoshop came out and everyone commented that images were ‘photoshopped’. In a few years we will look back and laugh when AI gen content is normal course.
English
The thing is that AI slop can have pretty strong metrics (viewability, ad density, etc.), which makes it harder to detect. Legacy verification tools are also not very good at identifying it.
Trishla Ostwal@trishlaostwal
NEWS: AI slop is fooling the ad verification tools built to catch it. AI sites are beating real publishers on the metrics advertisers trust and now make up to 2.4% of open web programmatic spend, with TAG's CEO saying billions of dollars could be at stake. adweek.it/4pFOomc
English
Great piece by Gareth Glaser on Google Buyer Direct, which allows buyers to connect w GAM inventory and bypass a DSP, as well as SSP. “GAM doesn’t need AdX; it only needs an agentic-like interface, with Buyer Direct under the hood.” @HatesAdtech #adtech
adexchanger.com/data-driven-th…
English
The whole internet took this badly acted joke video as indicative of something deep about modern society.
We don't even need AI slop.
Colin Wright@SwipeWright
I've noticed that women on the left increasingly insist that the men they date attend therapy. I think this is due to how thoroughly many left-wing women view nearly every interaction through a kind of "therapy ideology" with its own special language. Because of this, many find it difficult to communicate with men who have not been conditioned to see relationships through the same ideological lens. They therefore seek men who have either already adopted the language and worldview of therapy ideology or are (at least somewhat) willing to convert. "Do you go to therapy" is the left-wing female version of "Do you go to church?" on the right.
English
@tomfgoodwin I agree with this, but I will never abandon the scientific method in the name of “brand”
English
All the people who say brands are dead need to recognize that brands are actually absolutely everywhere. And it’s spreading.
MB@missedherbus
What’s wrong babe? You’ve barely touched your Minions salad from the diarrhea farm
English
Before you judge a man, ̶w̶a̶l̶k̶ ̶a̶ ̶m̶i̶l̶e̶ ̶i̶n̶ ̶h̶i̶s̶ ̶s̶h̶o̶e̶s̶ see what ads he's getting in his Instagram.
Español
AI slop as a category is a fascinating one, and I think a little too loose. There's a ton of procedurally generated content today -- summary websites, overviews, etc. For a data-driven industry, we seem awfully vulnerable to subjective descriptions that appeal to emotions rather than objective descriptions that correlate to outcomes. I think perfectly reasonable to demand an estimation of things like grade level, emotional content, copyright status, other contextual signals to determine whether or not you want to buy something, and author name is just one of those signals. An author named Claude is still an author, and just by virtue of being AI does not mean it is slop.
English
@HatesAdtech problem is that not all advertisers want to appear next to some content. Whether it’s news, ai slop etc. it’s more about control
English
@101Programmatic @oliviaakory "causal explanation" in marketing? what are these guys smoking? We don't even have causal explanation in _existence_
English
great insights. I think a lot of marketers fail even at level 1 - to use last-touch data, it has to be really clean. I've seen plenty of campaigns where ads weren't just viewable but carried a lot of attributed last-touch post-view conversions, and the real impact was close to zero.
English
Incredible read from BCG on what is necessary to enable agentic marketing.
"Most organizations’ measurement capabilities focus on campaign metrics, while a few have progressed to incrementality testing. Eventually, the best will adopt agentic measurement. The progression yields better tools and more data, but it also entails a fundamental shift in the questions that measurement seeks to answer."
More in 🧵
English
@emollick peter watts, blindsight, language without cognition is a topic (chinese box, lack of consciousness). If you haven't rad it yet, read it, and you're welcome.
English
I have read a lot of science fiction and nobody really anticipated AIs in the way we have them.
I mean, you can project ahead to Culture Minds or Golem XIV etc., but the current & immediate future, where statistical models of language do the things that they do, seems unexpected
English