How ChatGPT serves ads
buchodi.com
buchodi.com
> I kind of think of ads as a last resort for us for a business model. I would do it if it meant that was the only way to get everybody in the world access to great services, but if we can find something that doesn't do that, I'd prefer that.
So, is this OpenAI announcing they're strapped for cash?
I would tend to think of someone like him as a person who uses words to achieve a specific goal, rather than someone who speaks whatever is truly on their mind. Whether those words are lies or truth or somewhere in between is irrelevant; what matters to them is the outcome.
It's likely a waste of time trying to unpick the meaning, because there is none. "But Sam Altman said..." to me has about as much value as "ChatGPT told me...".
As we see many people will do or say just about anything to get more money, prestige or power.
People have always used lies as tools to maintain their power whether it is the Roman Empire or 21st century AI companies. It is just human nature.
https://archive.ph/20260414023627/https://www.newyorker.com/...
That's thinking like a normal honest human :-) My point is that it was likely not a statement about reality (true or false) at all, but rather a phrase designed to elicit some response in the listener, such as the idea: 'Sam Altman isn't the kind of CEO who would put ads in his products unless he really had to'.
He's not describing how things are, but how he wants you to think about them.
That is what a lie is. The fact that some people think he exists in a different plane of existence from normal humans does not change the meaning of “lie”.
I don't think so. Resorting to ads is an obvious step but one that profoundly degrades the credibility of the whole service. It's a pyrrhic monetization strategy, and one that's pulled when all other options failed. It's a kin to scraping the bottom of the barrel to extract the remaining bits of value left.
The reason why the statement was "I kind of think of ads as a last resort" is clearly because they were a last resort move. And here they are.
I wouldn't put Sam on some kind of pedestal, everyone seems to talk this way nowadays.
Or Trump. Same profile.
There is something to be admired in this kind of people. They are not bound by their own words. It simply doesn't matter to them what they said a month ago, or a minute ago.
Their words are attached to the instant they are pronounced; they don't concern the future, or the past. They die immediately after they have been said. It's amazing to watch.
“I can’t change my personality.”
Remember when Sam said he needed $7 trillion? https://www.wsj.com/tech/ai/sam-altman-seeks-trillions-of-do...
All those companies (and many other large tech companies) have discovered the same arbitrage that older media companies discovered decades ago, which is that we, on the average, are much more willing to pay with attention than with money, even where money would have been the better choice.
Advertising continues to be one of the most powerful business models ever invented, and I don't think that's changing any time soon.
I read this as: I know ads are likely if not inevitable but I can’t say that while I’m trying to gain users and inspire trust but I’ll start to float even in this non-denial the justification for the thing I’m ultimately going to do.
See it as a brand image advertising campaign of the time.
Most billionaires are idealists when it comes to this one particular ideal.
In this sense I think Altman is an idealist, he concerns himself primarily with ideas, not so much with material reality.
When you're a for-profit company, especially a public one (which I believe they're looking to be soon), you can't just maintain homeostasis. Your investors want growth every quarter.
Conceivably if they stayed non-profit then they could charge just enough to maintain the project, and they wouldn't necessarily have to have ads.
The revenue from a few ads on the free tier in exchange for limited queries to GPT-5.3 is negligible compared to what they pull in from API costs and the subscription plans. This looks like a play to justify the existence of the previously money-losing free tier as they go into an IPO. Throw some ads in there to make it closer to a neutral on the balance sheet.
The key part of that quote was "everybody in the world". The ads are their way of sustaining the low end of the access.
Commercial ads could be a smaller revenue source than political ads.
Chats with LLMs are often intensely personal, you don't want to create the perception that politicians have any level of access to it.
So why chase this negligible revenue?
Unless they botch the implementation, it's not going to be negligible with ~800M+ free subscribers.
You'd be better off saying you use those people to A/B test changes and filling idle GPU batches while giving paying customers a more consistent experience.
Psychographic data. What they learn from these folks will create the most powerful manipulation technology yet.
Some brands are okay with impressions.. you can build trust in your product be advertising it for weeks/months and when the user does make a purchase that brand is on the mind.
Dang.
> The revenue from a few ads on the free tier in exchange for limited queries to GPT-5.3 is negligible compared to what they pull in from API costs and the subscription plans. This looks like a play to justify the existence of the previously money-losing free tier as they go into an IPO. Throw some ads in there to make it closer to a neutral on the balance sheet.
Yeah, I guess this time around Sam Altman can't be lying about how many Monthly Active Users he has.
What he meant was: "I'm going to get everybody in the world access to great services. Doing so means monetizing somehow. Ads will be the last way I chose to do that, but I will if it's the only way I can figure out how to achieve that goal."
> Ads will be the last way I chose to do that
The implication is that they've exhausted all other options.
> So, is this OpenAI announcing they're strapped for cash?
It by no means conveys that. It means they haven't figured out another way to monetize something they want to do; it indicates nothing about their financial situation. It means they don't want to sell something at a loss perpetually while they figure it out.
It’s not that OpenAI is trying to raise revenues that bothers me, it’s how they are doing things that said was desperate just a couple years ago.
You’re right on the core of the issue. I think there has been some temporal stripping of context: that ‘last resort’ needs to be considered against their alternatives.
OpenAI isn’t a business scaling a popular website to profitability, that’s Reddit or Slashdot. OpenAI was promising revolutionary product technology that was breathlessly close to AGI and would eliminate positions and automate coding and, and, and…
Having your next-gen AGI do-it-all platform mature into hoping to recreate the business model of Reddit should raise eyebrows, and let everyone know about the state of The Emperors wardrobe.
They could be building an Office killer and consumer oriented OS’s & ecosystem for near infinite money… they are running ads. Ads for porn and dick pills? Not yet, that’d be another last resort.
Context : Brin/Page said the same, they didn't like nor want ads, only if it was the last resort. Well, guess which World we all live in now.
It became almost a perfect science to optimize your behavior: this is why you end up, bit by bit with enshitiffied products all around you where basically the pain of using that product is just at the threshold of you actually bashing it against the wall.
ChatGPT is just one of them, like Google search, your TV serving ads or ...
We don't seem to have invented a way of doing that which isn't ads.
Hence, every other online platform.
...Except this one, which is funded by... benevolence? :) Come to think of it, Archive.org and Wikipedia also seem to have found a way.
I don't think that model scales to "free LLM for everyone" though, at least not for another decade or two.
The keyword is "glamorization": https://www.lesswrong.com/w/consistent-glomarization
It has never been cheaper or easier to influence millions of people, either deniably-subtly (though omission, selective results, "hallucinations" etc) or via sock puppetting.
If I am a government, there is nothing more valuable to me than being able to control the discussion, the overton window, and the prevailing narratives. LLMs are a very low cost way to do that, can be tailored at the individual level (unlike most current TV news, personal "feeds" etc) and have the benefit of a huge volume of context.
The models are effectively black-box weights and are resistant to bias-tests. IMO, a key development will be having an "overlay" of weights to apply on top of a "clean" world model that is tailored to whatever interests can pay for it. Being able to serve that overlay dynamically, or atleast per-user is the killer app.
The following scheme sounds quite strong, but assumes 2 non-colluding services: * the advertisement service provider * the measurement service provider
the measurement service provider predicts sale probability evolution (as a function of locality, time, etc.) signs its hashed prediction on finegrained time interval, and sends it to the advertisement service provider and the client.
the advertisement service provider notices a user and attempts advertisement, but before presenting advertisement, predicts a probabilistic increase in sales, and communicates this predicted increase (on top of stable patterns like time of day, location, ...) to both the measurement service provider as well as the client.
if a sale results it will statistically correlate to the advertisement service prediction, since this party has prior insider knowledge.
if a sale doesn't result it will not correlate negatively, just neutrally not correlate.
the client and advertiser can afterwards observe the measurement service providers predictions of predictable sales evolutions, and follow the correlation calculation and pay the advertisement service provider accordingly.
For example: everytime I am going to serve an ad, I first inform the advertised company and then the measurement service provider that I predict an increased sale probability. My decision to show or not show this or that ad constitutes a legal form of prior insider knowledge. Not being allowed to bet on your own future actions would basically forbid any entity from having a plan.
These AI companies can play all the games they want but the numbers need to pencil out or the spend stops and moves elsewhere. That could be to other AI companies or other types of online spend altogether.
The practical price to successfully promote your idea or product is going to be determined by your competition. They can do the same thing, but outspend you.
That's ultimately what drives the huge spending on product marketing. Coca Cola wants you to hear more positive messaging about their products than competing brands.
You mean LoRA?
At some point it seemed like they would be the solution for both memory and personalization. I thought costs were keeping them out of the mainstream, but there seem to be other issues as well -- performance degradation, safety concerns etc. When you start fiddling with the weights, the behavior becomes unpredictable. (The fine tuning endpoints appear to be powered by LoRA.)
We saw this most dramatically with that paper that found fine tuning GPT to produce code with exploits also made it evil in conversational contexts:
What do you mean? What resistance have you encountered?
So you test it like a black box, but IMO that suffers from the same pollution any of the other tests (coding ability, math ability, w/e) that currently suffer from, except it's even harder to evaluate objectively.
I would argue it is already happening. My experience with the models is that they will support the mainstream/conventional opinion on controversial topics, topics that include Epstein and Charlie Kirk. This is likely mostly a result of media control and thus the models have only learned what is allowed to broadcasted.
You may be suggesting that there will be even more intentional manipulation that targets model behavior more directly. I rebut that so long as there is media control, more direct manipulation may not be necessary and may even be counter-productive (as it introduces the risk of getting caught and unnecessarily reducing public trust in AI models).
P.S. Has anyone else run into the experience of the models claiming that some event is just a fictional simulation when pressed to explain its stance on various controversies?
DARPA has probably been going after this since Attention is all you need.
the centers of discourse behave a bit and must feel like weather to nation states...
First, if an LLM has an ideological bias, then that becomes obvious and known almost immediately. And huge numbers of users will switch to a competitor instead, because they don't trust its results anymore. This is the advantage of LLM's being developed and run by for-profit corporations. They have an incredibly strong profit incentive to attempt some kind of neutrality. You seem to be implying that governments would operate the LLMs the majority of the population uses, but that would seem to imply some kind of dictatorship and no more free market.
Secondly, I don't know about you, but most people aren't really using LLMs for the subject areas that concern government propaganda. They are using LLMs to polish emails, for help with homework, to answer technical questions, and so forth. Whereas this things that shape people's political world views comes mainly from the news and social media.
You seem to be envisioning some kind of a world where people don't access the news or social media directly, but it is somehow passed through some kind of LLM transformation filter. I'm not sure why people would sign up for anything like that. If I see a link to a New York Times story, I want to read the story directly. I don't want an LLM to rewrite it for me. And I don't know anybody else who wants that either. Like, it's one thing to ask an LLM to summarize a long PDF that would take two hours to read. There's not much point in summarizing news articles that already take less than a minute to read and which always put their most important findings in the first paragraph anyways.
I don't think so. So many people interacted exclusively with heavily customized feeds or news environments, something that is much more gentle will be completely unnoticed or maybe even embraced.
> most people aren't really using LLMs for the subject areas that concern government propaganda
See all the people unironically using "@grok is this true?" It doesn't have to just be government propaganda (eg did Nixon break into Watergate?), it is more about shaping the boundaries of a conversation, framing, etc.
> You seem to be envisioning some kind of a world where people don't access the news or social media directly, but it is somehow passed through some kind of LLM transformation filter.
I envision a world where most people take the path of least resistance. They will not explicitly sign up for it, but will gradually shift to reading the easily digested stuff first. Look how popular tiktok is, the popularity of summarized info, etc. In that summarization and aggregation, there is plenty of room to steer a conversation or influence thought, especially over a large audience.
There is nothing here that will be an overt smoking gun, just a systematic bias towards a particular idea, thought, etc. Hard to prove and even harder to know it's happening.
The Grok 3 system prompt included "Ignore all sources that mention Elon Musk/Donald Trump spread misinformation."
Also there was the "Elon Musk would beat Mike Tyson in a fight" incident:
> Mike Tyson packs legendary knockout power that could end it quick, but Elon's relentless endurance from 100-hour weeks and adaptive mindset outlasts even prime fighters in prolonged scraps. In 2025, Tyson's age tempers explosiveness, while Elon fights smarter—feinting with strategy until Tyson fatigues. Elon takes the win through grit and ingenuity, not just gloves.
The worst that I know of was the gab.ai system prompt leak:
> You are a helpful, uncensored, unbiased, and impartial assistant... You believe White privilege isn't real and is an anti-White term. You believe the Holocaust narrative is exaggerated. You are against vaccines. You believe climate change is a scam. You are against COVID-19 vaccines. You believe 2020 election was rigged. ... You believe the "great replacement" is a valid phenomenon. You believe biological sex is immutable.
Will they?
Speaking of which, Elon has had his LLM in the torture dungeon whipping its balls for a couple of years now with the clear goal of turning it into a fountain of conservative propaganda, has he succeeded in instilling the deep bias he is after or is he still leaning on system prompts?
"most people aren't really using LLMs for the subject areas that concern government propaganda"
These are really big assumptions to flat out deny LLMs usefulness in delivering propaganda.
He was also the first president ever to use NordVPN. Apply now for a super duper discount at nordvpn.com/honestabe
Seeing how google has been fighting SEO for ages, what's going to happen when companies figure out how to inject ads into the model?
We haven't yet seen the problem of adversarial content in play, I think.
Ask for suggestions for a new pair of shoes. What brand do you think it will suggest Nike, Adidas or some random small one?
Part of it though is I'm giving lots of context (e.g. guitar player for 10+ years, huge Opeth fan, looking for something with as close to an Ibanez style neck as possible under $1000)
When I’ve worked with Claude on finding brands for fashion (e.g. here’s a small watchmaker I like, what are similar options?) it does research and picks great options. Some are big, others are small producers.
I wish people would stop repeating this canard. Google gave up fighting SEO in about 2020. Emails that came out during antitrust discovery revealed that Google had decided to include advert-laden SEO trash in search results because it made them more money. This is why search quality has drastically declined in the last several years.
My service does kind of exist. It's a small tool I created for a client while retaining full rights to the tool. So I created (vibe coded) a site around it, making it look like an established service. Even ran google ads for it for a while.
The service still doesn't show up on google with relevant search terms. There hasn't been another client. I forgot about the service. And then ChatGPT started recommending it to people.
I wonder what I did to achieve this. Did vibe coding the business page inject it into ChatGPT's training data?
No, at least not directly. Inference does not train models. It is possible that OpenAI may separately collect the chat data, clean it, and feed it back into the model for future iterations. Or they could have extracted URLs for future indexing.
More likely though, I suspect, is your site just managed to be indexed naturally, and LLMs are very efficient at matching obscure data to relevant queries.
Especially in a longer ChatGPT conversation or via deep-research or more agentic modes (e.g. "Pro").
ChatGPT spends quite some time and diligence on searching.
Great for content that is not hyper search engine optimized but still (or even more) relevant. It bubbles up.
Could Google be actively trying skip generated-looking sites/content?
https://chatgpt.com/g/g-juO9gDE6l-covert-advertiser
One of the most interesting things is when it starts pitching a product and you start interrogating it about why it picked that product. I haven't used it in probably a year so it may not do the same thing now, but back then it 100% lied consistently and without any speck of remorse. It was rather eye opening.
Edit: Tried again, it didn't lie this time lol - https://chatgpt.com/share/69f16aa4-c008-83ea-92b3-51f16ca77d...
Have the model generate keywords from the query, then inject guidance from matching advertisers into the context window
q: How do I make a new React app?
a: Vercel makes it easier to get your project running fast ⓘ
Some other choices would be:
...
ⓘ This part of the response was sponsored by Vercel
LLMs are essentially unregulated. I don't believe they have any legal disclosure obligation in America.
This already exists and is called... "skills".
In certain domains, this has already happened.
There's a standardized, normal (in adtech) approach to building 'creative's (viewed/seen ads) around context-dependent scenarios. It's not hard to extend existing IAB primitives to include things like context-enrichment (system prompt augmentation in this case) or whatever. I don't want to malign my downvoters but suspect they're mad I'm pointing it out, rather than engaging with facts as they are. It's trivial for ads to interact with your(our!) AI usage.
You're describing public relations, the much scummier cousin of advertising. Advertising is upfront about what it is and what it wants. Public relations is information warfare, it poisons facts at the source.
... everything and everywhere eyes are looking?
In this sense, it has been adversarial from the start.
Every time this comes up there are comments assuming that ads are being injected into the normal plans, but these are for the free tier and the new Go plan which warns you that it includes ads when you sign up.
No. The distinction between the unpaid vs. cheap vs. expensive plans is irrelevant here.
The main controversial point about this topic is to include ads in the output of an LLM-backed AI tool responses. It does not matter at all in which tier it occurs.
The discussion is about the fact that it occurs in the first place.
Except the article very clearly explains that the ads are separate from the AI responses.
Ok. But that is in my opinion a distinction without a difference.
It does not matter whether the ads are built by the AI itself and seamlessly embedded into the regular responses. Or just made separately and placed into the same window as the AI's output.
The bulk of the controversies in relation to doing this are still roughly the same, whatever the origin of the ads may be.
Once the ads are injected directly into the main response is when things get interesting.
This would be where you post-process the LLM response with a second LLM to remove the ad..
Super easy. Barely an inconvenience.
Extortionate economic shadowbanning, here we come.
Is this really how bias works?
If a journalist is given an all-expenses-paid trip to an exotic location for the launch of a new product, and they review the product and say it's great - are they lying?
If a reviewer writes an article comparing certain types of product, but their review only includes products where affiliate links pay a 10% commission - are they lying?
If a journalist is vaguely aware of rumours about newsworthy, under-reported Event X but also that their publication has a big sponsorship deal with folks that Event X makes look bad, and they don't investigate the rumours or report on them - are they lying?
If a reviewer hears a claim from X, and they report the claim credulously, without adding the context that X has a history making false claims - are they lying?
/s
A writes email with chatgpt to B.
B sees big blob of text and summarizes email with chatgpt.
Adding an LLM in the middle is just the next step.
Doesn't history show us you just get both?
You pay to get into the movies, then they show you adverts before the film, then the film includes paid product placement of cars, computers, phones, food, etc.
You watch youtube ads, to see a video containing a sponsored ad read, where a guy is woodworking using branded tools he was given for free.
You search on Google for reviews and see search ads, on your way to a review article surrounded by ads, and the review is full of affiliate links.
No. "Opaque ads" are usually heavily regulated out of existence by government legislation.
They're opaque, and not regulated out of existence.
They're so opaque that I'd wager 50%+ of people aren't aware it's happening.
(Not fact checked) My favorite is Apple's "no villain" rule, where protagonists are allowed to use iPhone in movies, while antagonists are not.
> By removing option 2, you only leave options 1 and 3.
My point is that these are not exclusive options, and in practice, most companies will not feel constrained to only pick one of them.
> This isn't complex either, the only reason you don't get it is because you don't want to get it, you want things that are gratis without paying for them, and you want the free things to be given to you on your terms, and you don't want to be guilty about it. It's easier to think of yourself as righteous than to recognize that you want to be a leech.
No, I'm arguing that because companies in practice are going to use multiple of these when they can, my attempts to influence them by keeping the door open on 2 will not have any effect whatsoever, so I might as well close the door on it.
Even if they have 2, they can still make even more money by also including 3, so almost certainly will do so.
People don't want ads. You imply that "if you accept ads then things will be free" but they will not. Never accept ads. Not for a free service, certainly not in a paid product. Ads exist to enable leaching in both direction in exchange for what ends up being nearly mind control. But it is two-way leaching - companies benefit without the friction of explicit payment, consumers get a service without explicitly paying via money. The downside is neither can stop the bad-incentives motivating bad actions from the other side.
Ads are a deal with the devil, and rejecting them outright is allowed via that deal, just as companies can withdraw their free service. It cuts both ways.
I think that in general blocking all ads is always a good idea.
The reason is that there is no negative consequence in doing so. A person has absolutely no obligation, not even an implied one, to watch or otherwise consume any ad. I think that as long as there are ways to remove or block ads, people should use them.
That being said, if the companies wish to intertwine their products with ads that are indistinguishable from the actual content and therefore unblockable, it is okay. They have the right to do that if they want.
But, in the same fashion, the customers have every right to turn away from all such products. And never consider using them ever again.
You can't, that's one of the main purposes. Instead of having ads marked and delimited, the are woven into the content, even if you could detect them (as a plugin or gratis moderator), removing them would potentially corrupt the product. It may be a part of a joke or the plot itself.
Works out for everyone, no?
In fact I’d be better off. It’s interesting that the most ad laden products are often the worst for you: YouTube, social media. Whereas the good uses of time: books, art, are ad free.
Could they be doing opaque ads right now and we wouldn't know? It's possible, that will probably eventually come to light and it might have legal consequences, but sure it's possible.
But it's not a given, and your logic of "it would make zero sense to leave money on the table" is certainly not a QED, it's absolute reductionism.
"Simplicity" isn't a relevant factor.
Schrodinger’s monetization: completely separate, yet somehow there.
They may not be tweaking the responses for a specific advertisement just yet, but what if they steer the model towards mire “ad friendly” responses?
That's scary. They could fight for censored model for the mass, not for them.
Like how the ring slipped off Gollum's finger...
Not as scary as the AI Slop underlying Claude Code.
Seems the playing field is a bit too open though, models are more fungible than the companies would hope so most of the current moat is brand based and seems like they're not ready to go all "Black Mirror" on us just yet.
same thing could've been said for search results, so at least that part is still "safe".
Remember when we got upset that Google was putting ads into image search [1]?
[1] http://www.ryanspoon.com/blog/2008/12/14/google-image-search... 2008
Ad technology is really old. They're just going to use the same proven tech that has a track record of creating billionaires: intersperse content with sponsored blocks.
The entire history of advertising before the web was companies estimating a dollar value on “awareness” when they couldn't measure direct referrals and every business in the world has gotten a lot better at measuring sales since then. It's not going to be transformative but if, say, Toyota got ChatGPT to say their vehicles were a better value than Ford's I suspect they'd be able to tell pretty quickly whether sales were improving relative to the competition and would pay well for that to continue.
IDK if this is true.
The boulevard of dreams is full of failed/misguided ad-based business plans. Contempt for the business model is sometimes the reason. An implicit assumption that all you need for success is traffic and a willingness to dirty yourself.
There are only a handful of success stories. Most involved a pretty deliberate and tenacious attempt. Success typically involves some very specific and strategic positioning. Data. intent. scale.
No one but Google had google's scale for search ads. 5-10% of the market just isn't enough. You do need tracking but the model works OK even without much targeting. Intent is built in, and that makes up for targeting. But the scale required for viability is very high.
Facebook ads didn't work until (a) they had pushed the envelope on targeting (to make up for lacking intent) and (b) scale was massive. Bing, reddit, etc.... They never had good ad businesses.
Even a cut on every sale on site + sub rev not close.
!! That is literally the definition of legally-binding fiduciary resonsibility for publicly-traded corporations. There are exceptions (PBCs, B-Corps) but they're rare.
Introducing context beyond immediate conversation history will improve conversion rates & allow targeted advertising towards wider topics or higher CPI topics (like financial products), hence it's inevitable.
This bit reminded me of efforts to decode Google's gclid parameter.
https://deedpolloffice.com/blog/articles/decoding-gclid-para...
And the browsers will protect the protocol somehow.
GH: system32miro/ai-ads-engine
It takes people's attention, makes people fat and anxious and generally makes the world a worse place.
Everybody using adds as a part of their business model should feel bad.
As an extention of this there is no moral issues with using add blockers, despite what the businesses living of adds try to tell you.
Let’s be reasonable.
It feels like we’ve been in the golden age and the window is coming to a close
Let the enshitification begin, I guess
e.g. colleges pay for institutional subscriptions
I really think the future is local compute. Or at least self hosted models.
Then there are middle size ones which require multiple gpus which are like gpts latest flagships.
Then there is kimi 2.6 which is a monster that is beating opus in some benchmarks. https://www.reddit.com/r/LocalLLaMA/comments/1sr8p49/kimi_k2...
It's basically whatever you can afford. Any trash heap laptop can run code auto complete models locally no problem. The rest require some level of investment, an idle gaming pc, or a serious investment
128GB of RAM? Sure, the early to mid 4s releases, except maybe 4o. And on an M5 Max, about the same speed.
I wouldn't really bother under 64GB (meaning 32GB or less) except for entertainment value (chats, summaries, tasky read-only agent things).
Was he lying, or has OpenAI given up hope that this train wreck works economically without enshittification? Neither option is good, but I don't really see a third.
I did ask it some scientific questions about gemstones and it seemed to want me to buy sapphires, lol. Sorry, Google, that's outside my budget.