AI + Ads

No Puns In This Title

THESIS: Every platform shift rewrites the ad playbook. Search created Google’s ad machine, social created influencers (as we know them today) and algorithmic targeting, and AI will do the same—but not by copying “four blue links.” Because LLMs trade on trust and authority, ads that feel intrusive won’t work in the same way. The real opportunity lies in new interfaces, such as AI agents that collapse friction in the hardest transactions, unlocking monetization models we haven’t seen before.

“The next Bill Gates will not build an operating system. The next Larry Page or Sergey Brin won’t make a search engine. And the next Mark Zuckerberg won’t create a social network. If you are copying these guys, you aren’t learning from them.” – Peter Thiel

Last week I wrote about how the business changes that seem obvious when it comes to platform shifts rarely end up being as straightforward as they appear at the time. You can’t just jam radio on the internet. That might work, but the real opportunity is generally doing something new that the platform shift allows for the first time ever.

One of the examples I briefly touched on what Ads. As a VC, it is hard not to salivate when thinking about ads + AI. Google is a Cash Machine, largely because it’s ad business is the best ever invented and the search product radically changed the way people interact with the internet. LLMs, in their current state, are search, but HYPERCHARGED. So it would follow that the ads built on AI should be EVEN BETTER than those found through Google.

But this assumes that our current state of LLMs and how humans interact with them is the end state. Which, seems unlikely. Even Sam Altman, who would benefit greatly (maybe) if the current interface and interaction with AI was locked into place, said “AI needs a new terminal because it changes the way it interacts with computers from the ground up”. The guy running OpenAI thinks that our current generally accepted terminal interface (meaning a screen where we input text and receive output) is not really even close to how people will be interacting with AI. It seems unlikely that it’s one big chatbot.

We are all basically Strong Bad

So it’s safe to assume that if AI / LLMs allow us to interact with the internet differently, and consumer behavior changes, then the way that we interact with Ads will likely also be different. It won’t just be “four blue links”.

Ads aren’t going anywhere anytime soon – the human attention is too important and ads are too effective. A lot of companies are even seeing current LLMs driving traffic and growth for them today. That’s not because of ads, but it does show the power of the product. Ads even exist outside of capitalist confines - Communist Russia had ads. Yes, they were state-controlled propaganda, but that’s just and ad for the communist party’s ideals. That’s a considerably worse, but still parallel, idea. So it’s safe to assume that ads are inevitable and that the interface will change, so ads are going to change as well.

As a case study, the rise of social media led to the rise of influencers. And while influencers are not necessarily brand new (i.e. Martha Stewart), their new prevalence because of internet distribution, the way brands interact with their customers has been updated. Companies have always had pitch people, hocking their wares for them. But “Internet Influencers” allowed for a more targeted approach, with (hopefully) a more attributable outcome (affiliate marketing, etc.).

To elaborate on the case study already referenced, Search allowed for a new ad paradigm as well. Google is the most profitable company on the planet (not hyperbole) because it revolutionized ads. It’s not just adwords and Google’s reverse auction, but the nature of ads in the search experience in general. Sometimes, someone is searching for something and all they want to see a product they can buy. They google “diapers” and want to buy a diaper brand. Getting an ad for diapers is actually helpful. The experience of the ad is a net positive for the consumer. So not only is the consumer giving up attention – which is worth something in and of itself – but it could also be converted to a buyer, which is very clearly worth something.

Then social also made ads a borderline enjoyable part of their feeds. And improved algorithms made those ads more targeted / personalized based on your social experience. In this sense, AI has already changed ads pretty drastically. Social media is not about your own personal social graph anymore. In fact, current “social media” is not social at all. It’s entirely driven by an algorithm to serve a consumer a facsimile of their social desires: groups they want to emulate, hobbies they want to pursue, trends that bring them joy, and BRANDS that could help them. The algo(s) made an inefficient market (word-of-mouth from people you know) and made it much more efficient (people you are mathematically more likely to be influenced by). That process is all driven by machine learning & AI built into a social platform’s algorithm. Your ads aren’t targeted based solely on who you are (demographics), but also because of the actions you take.

Now that idea might make your skin crawl. There are people who would tell you that ads are bad for society. That’s BS. Ads are a net positive for society. Yes there are bad things about ads, but ultimately, they have always been about consumer education. Good guardrails have made them effective tools to make our lives better. If algo ads make you upset, be upset with the platform, not the technology.

Machine Learning impacted ads, not because of the technology being so radically new and inserted into an existing platform. ML made products different (social feeds, FYPs, forever scrolling, etc.) and this allows Ads to be built in a different way. The product wasn’t built for ads, ads were built for the product. Behavior changed and ads simply followed.

So when behavior changes because of AI, we need to expect our ad experience to change as well. I don’t expect a lot of banner ads in my GPT terminal. Nor do I think it will look similar to Google’s Four Blue Links (ad headlines). Additionally, If I ask my AI a question and the response only shows up in a GPT response because someone paid for it, that somewhat defeats the purpose of AI / LLMs in their current context. Unlike current search products that are simply trying to put you in the direction of an answer and allowing you to discern that answer’s feasibility on your own, LLMs are just trying to give you THE RIGHT ANSWER. That’s why so many folks are up-in-arms about AI hallucinations. Current LLMs provide a deeper level of authority, or at least, that’s the goal. That is a feature, not a bug.

As a result, LLMs are actively trying to build trust among users. Having ads baked into the experience can erode that trust. How do I know that my deep research of “the best car for me to buy based on xyz parameters” will have an accurate output if a car company is paying to be inserted into the answer? There will be some level of tolerance for that, but I think the LLM market is too competitive today to test that tolerance too much. Even some of the LLM tools that are more ads friendly have pulled back on their ads business, such as Perplexity. Maybe they are pulling back on their ads business because it is not as a successful as they would want it to be. But even that tells you something – we haven’t figured out how ads are going to work on this new platform yet.

User trust remains relatively low - hallucinations are still at the top of people’s minds. And a lot of investment has been made into this space - meaning high capital returns are expected. So AI companies are going to be focused on two things: building trust and figuring out monetization avenues. And if traditional ad playbooks erode trust, leadership at AI companies are going to have to get creative and find new revenue paths.

Again, it seems unlikely that the terminal view of AI that we current operate through does not seem like a solid end-state. But one thing that seems likely (and brand new) is that AI results in more agents, doing stuff for us, including deep research. So maybe we should expect ad / monetization platforms that are more tailored towards agents. For instance, when an agent finds a product that can help us based on a prompt/interaction, it might be empowered to purchase that product for us. It’s already done the research for us, now, can it plug into the acquisition stream and make our lives all that more easier by actually transacting.

Now, maybe they can do this for diapers. Speaking from experience buying diapers is slightly harder than I expected it would be before I had kids. But, thanks to Amazon and other ecommerce giants, it’s a pretty seamless experience overall. So there might not be an incentive for those players integrate heavily or give a big cut of transaction volume to the AI companies. But what about things that are hard to transact upon? Like insurance, a new car, a house, a health payment, a spool of copper wire, a replacement part for a John Deere tractor, etc. Affiliate link marketing rarely extends to more complicated or regulated markets like these.

I am not sure the infrastructure really exists for that yet (although I am sure someone is working on it!), but I would expect a big AI company to try and benefit from those transactions. And the more complicated the transaction, the more equipped they are to demonstrate value (or whoever the agent builder is). And it’s not just about pure transactions, but events that lead to transactions. What if I am looking for a therapist that my insurance covers, that also does telehealth visits and focuses on mental health issues that are pertinent to me? That’s a ton of work a person can do, or a pretty quick task an agent can take on. Could that agent take a finders fee, regardless of who gets selected?

That might be a stretch, but that’s sort of the point. We don’t know the best way to monetize this stuff, but I anticipate it is going to open new doors we haven’t thought of before. For Google, ads improved the product - made the consumer experience better and more insightful. What is the ad paradigm that will make AI products better?