Rob's Notes 5: AI Will Shrink the Size of the Advertising Market

“I don’t listen to sales pitches. Talk to my software assistant.”

tldr; AI has the potential to revolutionize the advertising market by making it more personalized, efficient, and data-driven. This could lead to the reduction or even disappearance of many traditional ads.

In 2016, I wrote that AI would dramatically shrink the size of the advertising market over the next 20 years. With recent advances catalyzed by the public availability of large language models and generative AI that we’ve seen over the last 6 months, I feel more strongly than ever that ads will fundamentally change (and many will disappear) as a result of AI. 

If you’ve used any generative AI tools, you (and your employer!) might have had some trepidation about how the information you type into such tools might be used - you may question if what you type is getting you a better answer, making the company providing the tool smarter, or both. To overcome this trepidation might require having AIs on both ends of a transaction or query.

We will need help building and deploying them, but it is increasingly likely soon many of us will have AI agents we privilege with enhanced access to our messages, personal data, possessions, needs, and desires, and we’ll empower these agents to ‘negotiate’ or entertain offers on our behalf with sellers. These tools will need to include privacy-preserving technologies and abuse protections to prevent firms mining data from our user agents via various adversarial strategies, but the rise of consumer agent AIs seems inevitable now.   

Imagine you and your partner got a free beach vacation from your personal AI listening to a timeshare pitch! Unlikely perhaps, but highly desirable users’ agents could get small payments for entertaining advertiser pitches. New fees or bonuses from these models likely won’t be enough to replace the ad revenue lost as we reduce the waste from today’s uncertainty-weighted online ad markets. While I don’t expect AI to necessarily reduce the number of injury attorney billboards I’m going to “incidentally” see while driving on the freeway, online advertising is already 3x the size of the TV ad market¹ so the absolute impact will be large.

Here’s what I wrote on Medium in June 2016 (lightly edited²): 

“The advertising market will shrink in size dramatically in the next 20 years, and one key driver for this change will be improvements in artificial intelligence.

Advertising is an incidental (often interruptive) product and disliked by many people already. Because data about the features and suitability of products and services to individual users and use cases is poorly structured, we rely on chance encounters with biased information supplied by advertisers that often appeals to emotion, instead of common sense. 

But this will change, and that data will inevitably become better organized. And as we then give smart bits of software (who will often have a chat-like, voice or email interface and be our “assistants”) access to our calendars, tastes, information about what we already own, our goals and things we need, and other personal information, we’ll establish a profile under our own control (not distributed amongst thousands of companies all relying on outdated information, trying to infer our wants and needs).

This will create a major wall between us and product sellers.

Consumers blocking irrelevant and annoying ads, or email spam filters, were but very small early tastes of this. Advertisers will have to talk to our AI assistant gatekeepers first before getting access to us. And that ‘wall’ will learn, become smarter and more data-driven over time, and share algorithms and insights with other bits of software (hopefully while keeping our personal data secure, though of course in this we have learned over the last few years, there can be no guarantees). Display advertising will become obsolete: fewer flashing ‘punch-the-monkey’ images trying to grab your attention incidentally on a page of content, and more data APIs and ‘preference feeds’.

There will be a lot more software AI “bots” which will probably have a lower tolerance for deceptive practices and will learn based on the ongoing feedback we provide to them (and will learn some fractional amount based on what other users are telling their software ‘cousins’ filling similar roles).

The future is about filters, and though ad blocking and spam filters might be where it began, artificially intelligent software agents and AI bots are where it’s going.”

Source: Midjourney

A few final notes

Ads have been a way to fund the provision of products for free or at limited cost to a broad range of users. Naturally there will be equity and affordability concerns as the ad market weakens: certain classes of users who cannot afford higher-quality AI agents could be taken advantage of and lose out (government or nonprofits could certainly provide access to some level of AI agents for users who cannot afford them).

These changes won’t be immediate and we’ll first go through a stage where text and images from generative AI models are used to expand today’s ad content variations for advertisers, or even generate more personalized ads on the fly.

Some companies monetizing via ads today may see all of this and decide to avoid these ad- or offer-based models entirely, and make their products pay-per-use or subscription, further reducing “free” options. And shrinking the ad market further.

1. Wall Street Journal, April 2023

2. Previous Article on Medium, June 2016