# Is AI already good enough to create the missing online advertising boom? > Source: > Published: 2026-08-25 00:00:00+00:00 **previously:** [Prediction from the AI-max timeline: Vibeshop](https://blog.zgp.org/vibeshop/) Online advertising is still stuck in its own version of the [productivity paradox](https://en.wikipedia.org/wiki/Productivity_paradox) faced by pre-Internet IT. Back then, Robert Solow said, “You can see the computer age everywhere but in the productivity statistics.” And now, Michael Farmer is pointing out that today’s surveillance advertising is [Ineffective for Many Advertisers and Unhealthy for Democracy](https://michaelfarmer.substack.com/p/ad-tech-which-creates-audience-balkanization). Analyses of the revenue performance of the top 60 advertisers between 2009 and 2024 show that 40 advertisers, or 2/3rds of this group, all of them major programmatic users, saw their growth rates fall below nominal GDP growth rates from 2009-2024. While the surveillance ad [duopoly](https://blog.zgp.org/living-with-a-bigger-ad-duopoly/) grows, the brands that *use* their advertising, not so much. And the squeeze is getting worse, not better. [Meta’s revenue is up 28% while claimed conversions on Facebook are up 15.7%](https://www.adweek.com/media/metas-tepid-revenue-outlook-undercuts-its-ai-spending-spree/)—the [customer acquisition cost squeeze](https://blog.zgp.org/living-with-a-bigger-ad-duopoly/) continues. While advertisers, on average, are doing worse, each individual advertiser’s dashboard makes it look like their ad spend with the duopoly is a win. Because the duopoly controls the measurements. In [Why Social Media’s Ad Dominance Is A Billion-Dollar Measurement Illusion](https://www.adexchanger.com/content-studio/why-social-medias-ad-dominance-is-a-billion-dollar-measurement-illusion/), Vicky Chang writes, Meta’s attribution infrastructure is clean, fast and easy to defend in a spreadsheet. A brand CMO can walk into a board meeting with an ROAS number, a cost per acquisition and a clear story. That’s not a small thing. Internal budget decisions rarely go to the channel that works best. They instead go to the channel that can be proven most clearly. But much of it is an illusion. When a platform owns the measurement, it is more likely to push the narrative that the ads are working. The measurement models make that easy to do. (And yes, this is the same kind of racket that the duopoly companies are trying to push into the browser with the [attribution cartel](https://blog.zgp.org/the-advertising-cartel-coming-to-your-web-browser/) scheme at W3C.protip: if you’re running an industry organization, have an up-to-date antitrust policy and enforce it, or big companies will turn your meetings into a crime clubhouse.) So if Google and Meta control the ad measurements that your boss sees, it’s game over, right? Staying in marketing means sending money to the companies that [push the next measles epidemic or dictatorship or whatever](https://www.abc.net.au/news/2026-08-06/ragebait-how-facebook-is-paying-controversial-creators/106940696), and there goes the chance to support anything win-win? Brian Jacobs writes, in [Lessons from Spock](https://www.bjanda.com/blog/lessons-from-spock/), If the professionals know this is crazy, and if they are as well informed as they are, why do their clients do the exact opposite to what logic suggests they should do? Are agencies these days so removed from the real decision makers in client companies that nobody with any real say over budget allocation cares about, or even hears their advice? Or are agencies so complicit in the hopeless placing of their clients’ ads that they would rather keep quiet? But what if that dismal, centrally-planned future doesn’t happen, and markets can win? Meta and Google were early adopters of machine learning, for their own villainous ends, but other AI projects, including [“open” LLMs](https://newsletter.semianalysis.com/p/are-open-models-catching-up), are catching up. Rick Bruner suggests borrowing some ML techniques from Big Tech and running them under control of the individual brand, in [RCTs Aren’t Just Backward-Looking. They’re Training Data for Optimization Models](https://www.linkedin.com/pulse/rcts-arent-just-backward-looking-theyre-training-data-rick-bruner-egmmc/) Imagine instead a platform offering RCT incrementality testing transparently to advertisers while accumulating those experiments into its own causal benchmark. Advertisers get causal proof; the platform builds a valuable knowledge base; and the evidence improves its models and optimization systems. With that kind of project in place, advertising decision-makers would have an alternative to those slick Google and Meta measurements that back up the decision to dump more ad money into Google and Meta. A marketer who wants to branch off from the [dystopia timeline](https://rjionline.org/news/big-techs-economic-takeover-can-be-beat/) can vibe code a measurement tool that could show how feeding money and customer data to the ad duopoly has lower ROAS than the positive-sum options. I don’t think that kind of shift would happen on its own, though. Marketing decision-makers will need some kind of crisis to not waste. More later. [The Multi-App Strategy That Helped Me Earn More Without Working Longer](https://therideshareguy.com/the-multi-app-strategy-that-helped-me-earn-more-without-working-longer/) by Sergio Avedian. (More platform choices means more money.) [Trump tried to curb clean energy. It’s booming anyway.](https://arstechnica.com/science/2026/08/trump-tried-to-curb-clean-energy-its-booming-anyway/) by Martha Muir. Producers can break even by selling solar and wind power for as little as $38 and $37 per megawatt-hour, respectively, compared to at least $48 per megawatt-hour for gas, according to investment bank Lazard. [Daily Reminder To Not Listen To Google’s AI Overview](https://whatever.scalzi.com/2026/08/19/daily-reminder-to-not-listen-to-googles-ai-overview/) by Athena Scalzi. Please please please stop Googling things, looking at the AI Overview, and assuming it has given you the correct answer to your query. Don’t trust Google AI with anything, even with inconsequential things! Because it will, and DOES, lie to you. (You can still [fix Google Search](https://blog.zgp.org/fix-google-search/) to remove the slop.)