"AI, no ads please": 4 words to wipe out $1tn
Introducing a series on AI vs. ads
tldr: AI poses a huge threat to ad-based platforms by slashing how many ads we see
The AI vs. ads series:
“AI, no ads please": 4 words to wipe out $1tn (i.e. this article, below 👇!)
So far, most of what I’ve read about AI and ads talks about the impact on the demand side.
For instance, buyers of ads will find it cheaper and easier to make creative with AI, which will inevitably cause job losses. Demand will increase because AI will help people target ads better.
Almost no one is talking about the massive threat AI poses to the supply side of the online ads business over the next 10 years.
Google, Facebook and TikTok make $383bn in ad revenue between them (2023). The whole adtech industry is worth $1 trillion a year. But four words could wipe this revenue out:
“AI, no ads please.”
(Three, if you’re rude to your AI.)
It’s the inventory, stupid
Availability of inventory underpins the entire adtech industry: the amount of advertising space platforms have for sale.
What is this inventory? Nothing more than our attention, sold by Big Tech via the pixels and speakers of our devices.
The size of this inventory, and therefore the $1 trillion figure, is directly related to how much time all 8 billion humans give to ads every day.
How do 8 billion humans consume so many ads every day that we create inventory for a $1 trillion industry?
Well, people want great content. So the platforms that provide it make sure that ads are positioned between, around, before, after and inside the great content. You’re going to see ads whether you like it or not.
But that’s strange, when you think about it. It’s a wholesale infiltration. Most of the content we consume is via our own devices. We own the pixels and speakers of our devices (even if Apple disagrees). Why are they betraying us — allowing Big Tech to use them to position ads wherever they like?
It’s complicated. The betrayal is propped up by many pillars: technical complexity, a lack of regulation, interoperability and therefore competition, Big Tech’s lawyers, the surveillance capitalism business model, and a huge gap in resources between ad platforms and those opposed to them.
Now AI is making those pillars unstable. AI has the potential to undo the betrayal and lay waste to ad inventory. And before you roll your eyes, I’m not spinning a David vs. Goliath yarn. This is not about underdog developers creating the next generation of AI-powered adblockers.
This is more like Goliath vs. Goliath: Apple, Microsoft, and OpenAI on the one side, Meta and Alphabet on the other.
Huh, really?
As I’ve started diving into this, many questions have come up. So many questions that I’m listing them below to not go crazy. Because the AI vs. ads scenario is so complex, I’ve decided to write about it in a series of articles, instead of addressing all questions at once.
But I’ll answer just one question right away: huh, really?
Will AI really destroy a $1tn industry?
Not necessarily. Plenty of things could stop this scenario playing out. But I do think it’s plausible. And even just a plausible AI vs. ads scenario is a huge threat to Meta and Alphabet, both riding high at (or near) their record valuations. It’s a scenario that, even in its weakest form, could wipe out $100bns of their value. Even if it’s just 1% likely — it’s still worth thinking about.
But I think it’s much more likely than that. And as a completely impartial third party, I’m happy to have a crack at that thinking.
That’s what I’ll be doing in the coming weeks and months. Make sure you subscribe to catch the whole AI vs. ads series:
Questions, questions, questions
Here are the questions I mentioned. I’ve gone ahead and ad libbed a few answers as starting points, but I’ll be diving far deeper. Your thoughts are very welcome.
How will AI reduce the ads we see?
By giving us the content we want, and nothing more. This is already happening — look at ChatGPT and Arc. But the exact mechanics need spelling out.
Which actual AI feature or product will reduce the ads we see?
A few, but AI operating systems (OSes) in particular.
What is an AI OS?
I’ll be defining this more precisely in the series, but something like:
An OS which uses AI to enhance existing human inputs (e.g. voice), and which allows those inputs to control some/all software installed on the OS.
Why not just talk about chatbots, agents, etc.?
AI will be a primary way we’re going to operate computer systems. Hence, it makes sense to talk about AI OSes. It feels like the natural unit of analysis to consider the primary role that AI will have in computing’s future.
Aren’t AI OSes going to cause the paperclipalypse or something even worse?
Maybe, but that’s out of scope for this series. Humanity going extinct isn’t that big of a deal compared to Meta experiencing a decline in its revenue.
Any more questions about AI OSes?
Hm, just a few. How do AI OSes relate to a chatbot like ChatGPT, or a device like the Humane AI Pin? Which existing technologies already look like AI OSes? (Quite a few, not least ChatGPT and Arc.)
What would an AI OS look like precisely? How will we interact with AI OSes? What is the mechanism by which human intention turns into action in existing OSes, and how does this compare to AI OSes?
What are the ways AI OSes will reduce ad consumption? Will ad reduction be a deliberate intention of the user, or just a side-product of how AI OSes give information?
What is nudgeware, and how do AI OSes link to it?
Will AI OSes be able to control all the software on your device? If so, how? Isn’t there a ton of thinking on this already? (Yes, and I’m reading through as much as I can find.)
Who will be the winners and the losers of AI vs. ads?
Winners: hopefully the consumer, as well as OpenAI, Microsoft, and maybe Apple.
Losers: Meta, Alphabet, TikTok.
Why would OpenAI, Microsoft and Apple want to reduce the ads we see?
They won’t, necessarily. But OpenAI have established a model where people are happy to pay a monthly subscription to get precisely the content they want, and nothing more. That paradigm shift in consumers being socialized to pay for information directly (rather than through ads) is a big part of the danger for Meta and Alphabet. More thinking through the plans and incentives of these players is needed.
Does the AI vs. ads scenario depend on the actions of OpenAI, Microsoft and Apple?
No, I don’t think so. I believe there will be a race to create the best AI OS, and in most versions of that race, ads lose.
How is AI making those pillars of the ad industry you talked about unstable?
That’s a big one, related to a bunch of questions above, and I’m not going to try answer it here.
How does AI relate to other sources of pillar instability, e.g. regulation?
C’mon, I have plans this evening. Stop it with this stuff.
How is AI already affecting the ad industry inventory?
A whole lot. Think of all those ChatGPT queries that are robbing Google of search ad revenue, as people realize they can get the information they want more efficiently than looking at search results.
So why is Google’s ad revenue doing so well?
Huh, I’ll look into it. Probably they just made a mistake in their numbers.
Which signals would we need to see for this AI vs. ads scenario to be coming true?
I’ll be figuring that out (and perhaps making predictions on them using Metaculus), but at a minimum, Meta and Alphabet’s ad revenue decreasing would be among them.
Why does interoperability matter to the AI vs. ads scenario?
Because it might affect whether AI will be able to get users exactly the content they want, and nothing more.
AI will need to somehow interoperate with the platform or website where the content was published in order to bring it across in any form, and the people running those platforms and websites are not going to be happy, at all. Expect lawyers.
That’s all for now. Excited to get started!
I see fewer ads as an absolute win. We left TV/cable for the internet to escape ads, we'll leave the internet to escape them again.
1trillion is roughly $2,500 per US resident. Let's just focus on the $383 billion from Google, Facebook and TikTok. That's still about $1000 (presumably per year) per US resident. First, I'm wondering what fraction of my income is flowing towards this business already. But second, you can be sure I'm not going to be willing to start paying $80 per month for AI services, no matter how good they are.
Then, on the other hand, you have a point. If we all start paying $20/mo, it could make a HUGE dent on the industry...