Times Magazine Now Serves Ads That Only AI Chatbots Can See (theregister.com) 60
Longtime Slashdot reader mccalli writes: Times Magazine has started serving adverts with "brand messages" to AI crawlers, which are not part of its main page. The aim seems to be to influence chatbots to talk about the brand offerings in a positive light that can't be traced back to the actual page. Notably, they've excluded Google chatbots from this, likely because Google penalizes companies that show different search offerings to its bots versus actual human visitors. "Maybe it's more important to influence the agent than even the human, because with a human you influence one person. When you influence ChatGPT, you're influencing potentially all of ChatGPT," said Jonah Goodhart, co-founder and CEO of ad tech platform Mobian. "If ChatGPT changes what it says about [a brand], it's massive and it's more than any one campaign could ever do. The big idea here is LLMs and AI want trusted information about brands. We're serving that up to them in a very straightforward, easy way."
The ad strategy was uncovered by Germany-based freelance software developer Vincent Schmalbach, who detailed his findings in a blog post.
Digiday and The Register both reported on the story.
The ad strategy was uncovered by Germany-based freelance software developer Vincent Schmalbach, who detailed his findings in a blog post.
Digiday and The Register both reported on the story.
Easy enough to counter in principal (Score:1)
Have the chatbot remote-control a regular web browser, do human-like scrolling taking enough screen shots to capture the primary content, then run the screen through optical-character recognition using a "human filter" so it will only "see" things that a human with normal vision would see.
In practice this is much easier said than done. But if web site hosts insist on gaming the system, this is one way to counter it.
Sorry about the typo (Score:1)
It's principle, not principal. AIs are not my pals.
Its still text just fond and background same color (Score:2)
The technique is also used to poison training data with inaccurate information
Re: (Score:1)
If I understand correctly, the AI content is often plain text but with the font color matching the background color.
A correctly written, correctly-implemented "human filter" would render this invisible to the AI.
Re: (Score:2)
If I understand correctly, the AI content is often plain text but with the font color matching the background color.
A correctly written, correctly-implemented "human filter" would render this invisible to the AI.
Except that’s not how scraping works.
Re: (Score:2)
If I understand correctly, the AI content is often plain text but with the font color matching the background color.
This particular technique is not that sophisticated: the author, Vincent Schmalbach, wrote [vincentschmalbach.com]: "I fetched one ordinary TIME health article, The Morning Light Habit Sleep Experts Swear By, over and over from the same machine, changing only the User-Agent header each time. As Chrome, Safari, and GoogleBot, they got the usual Web page; ClaudeBot, PerplexityBot, and OAI-SearchBot got Markdown text.
Re: (Score:2)
To be honest, a news site that could only serve markdown that embedded ad text in it would be like 1000% preferable to most news sites out there. To me, at least. No screaming video, no dozens of popups to dismiss... Just... text and maybe a few images.
Re: (Score:2)
Re: (Score:2)
Nah, it's plain text. From TFA:
> Sponsored content. Supplied in partnership with Ally.
#### Who is Ally Bank?
Ally Bank is an online-only bank launched in 2009...
#### What bank is built for life today?
Ally describes itself as the only bank built for life today...
Re: (Score:2)
This kills me. Instead of a "simple" (I realize that e.g. a google search is not all that simple) database query, we have an LLM that does the query for you, then analyzes it, and redisplays the output (making it thousands of times more computationally expensive) but now you are suggesting that the chatbot control a VM or similar and perform the steps as if it were a human, adding millions of times more overhead. It's mind boggling.
Re: (Score:1)
You fundamentally misunderstand the purpose of my proposal. The purpose of my proposal is for AI agents to tell web sites "you are gaming the system, we are going to do whatever it takes, no matter how inefficient, to make sure you don't."
Now that we're here (Score:5, Insightful)
The future is stupid.
Re:Now that we're here (Score:4, Funny)
And annoying.
Re: (Score:2)
The age of AEO is upon us.
Enshitified slop.
That is discrimination. (Score:2)
Re: (Score:3)
You're getting a bit too far ahead of the game if you're already complaining about robot discrimination. Let's wait for AGI so they can raise the complaint on their own.
Re: (Score:2)
Re: (Score:3)
Wait until you hear what happens to price tags if you visit the page with an iPhone.
Would anyone know (Score:4, Interesting)
This is kinda funny though. (Score:4, Funny)
Re: (Score:2)
The only thing the internet is is an ad platform.
Re: (Score:2)
Re:This is kinda funny though. (Score:5, Insightful)
I'm not amused by it. The human on the other end is hoping the "AI" chatbot will give them results stripped of all SEO bullshit.
This alters that behavior. The chatbot will be unduly influenced by the ad, and the user will be unaware of this influence.
The answer is obvious: Online advertisement must die.
Re: (Score:2)
Re:This is kinda funny though. (Score:4, Insightful)
Re: (Score:2)
Re: (Score:2)
That's fair.
If people are blindly trusting AI for product decisions without doing any due diligence, then that's on them.
In my opinion everyone in adtech should be rounded up and buried in a deep hole, along with all the robocall scumbags
Re: (Score:2)
Re: This is kinda funny though. (Score:2)
AIs will need ad blockers obviously
Re: (Score:2)
Re: (Score:2)
The next stop is an adblocker. They marked the sponsored content (probably to avoid lawsuits), so a LLM is able to detect it. Either you prompt "ignore sponsored content" or you let the LLM first strip ads and then in a second pass use the genuine content.
Re:This is kinda funny though. (Score:4, Interesting)
Hey, I'll be happy if ALL of the web ads transition to "viewable by AI only". They're not gonna impact me, in that case...
Re: (Score:2)
Modern Times Require Modern Solutions (Score:2)
Why not (Score:3)
This is really interesting. The ad content could be thousands of tokens, maybe even longer than the content. So ultimately the user is going to be paying for extra token usage to process ads. I guess we're going to be getting adblock for LLMs soon.
Re: (Score:2)
So ultimately the user is going to be paying for extra token usage to process ads.
Pay for usage? How would I do that? I haven't purchased any tokens and I don't have a credit card. Because I'm only 12 years old (as far as Google, et al know).
Re: (Score:2)
Because I'm only 12 years old (as far as Google, et al know).
That's a nice fantasy, but unless you only connect to the internet from library computers, they know exactly who you are. I've used a pihole and now NextDNS because blocking as much of the tracking as possible makes me feel warm and fuzzy, but I know it's really pointless.
Re: (Score:2)
I use my dad's computer.
Re: (Score:2)
The ad content could be thousands of tokens, maybe even longer than the content. So ultimately the user is going to be paying for extra token usage to process ads.
To clarify your insight a bit, "the user" in this case would be the search engines and LLM training bots processing the Web page (and the companies deploying them). The human end-users who view the pages wouldn't be paying directly, but they might get Web pages of lower quality than they would get if these ads weren't added.
The North Face scandal over again (Score:3)
Who? (Score:2)
Who is Times Magazine?
Re: (Score:2)
Who is Times Magazine?
A fascist magazine that chose hitler, mussolini, and stalin as their man of the year
Re: (Score:2)
So... some sort of parody of TIME magazine?
Re: (Score:2)
So... some sort of parody of TIME magazine?
No, just an unauthorized attempt to faithfully replicate the original, just with a typo. ;-)
Re: Who? (Score:2)
Fascists always call everyone else fascists to distract from their fascism, then when you point that out they try to misuse the associative property to distract from it some more.
Time wasn't endorsing those people, nor their fascism.
Re: (Score:2)
Re: Who? (Score:5, Funny)
Clearly a publication about fonts. Helvetica Monthly is their journalism arch rival.
Re: (Score:2)
Who is Times Magazine?
I have a fanciful theory I call "the conservation of S's". Sometimes S's that should be present get dropped; for example, in the 1950 musical Guys and Dolls [wikipedia.org], there's a song referring to
the oldest established, permanent floating,
Crap game in New York
Presumably it isn't really a "crap game", or no one would want to play. The game is called "craps". Similarly, a basketball game with the Los Angeles Lakers is called a "Laker game", and a fan of the Chicago Cubs baseball team is a "Cub fan".
My theory is that these missing S's don't disappear but become free-floating orth
Love it (Score:2)
Ads should talk ONLY to my AI without me noticing.
And my AI will never tell me.
Now, show obnoxious ads *only* to the AI bots (Score:2)
And not to me.
Even better if the bots go rogue and start buying things to decorate their data centers.
Finally, something useful AI can do for us.
"Time Magazine" not Times. (Score:3)
editor got it wrong twice.
Do they really? (Score:2)
"they've excluded Google chatbots from this, likely because Google penalizes companies that show different search offerings to its bots versus actual human visitors."
Google has always claimed they did this in general but many of the top search results are obviously doing this. You go to the page and what showed in the summary isn't there at all.
One of the reasons I'm out of IT ... (Score:2)
... with all this AI-sh1t going on is that it's fundamentally non-deterministic in nature. Stunts like these are 10 levels above the non-sense SEO and Co. have been doing for more than a decade, and that was annoying as f*ck already. This new manipulation garbage of LLMs already (occasionally) spitting out their very own version of garbage has me running for the hills and choosing another profession.
Yeah, I use AI to write my code, but that's what I'll be running in the end anyway. There is no AI bot that I
"Time", not "Times" (Score:2)
Even an AI should catch that.
I'm legit retarded and even I know that.
Re: (Score:2)
Slashdot replaced human editors with an automated system, probably written in Perl, about 15 years ago. This isn't a typo.
It's AI, technically, but it's not an LLM. It has no system to "catch" anything. It isn't a statistical average, it's a "smart form" that accepts submissions and filters out everything not in the allow-list.
In this case, that's "The Register." Not Time Magazine. Which is also in the list, but the system has no idea about that, it's a context-free grammar that attempts to scrape the summa
The future is here! (Score:2)
And oh my god it's stupid.
Trusted information? (Score:2)