Marketers Are Throwing Money at AI Discovery—And Flying Blind on Results
Let’s get something straight: the AI hype train is barreling through the media landscape at bullet-train speed, and marketers are shoveling cash into the engine without a working speedometer. The latest IAB survey just confirmed what anyone actually running campaigns already knows: brands are spending more to chase algorithmic discovery, but their ability to actually measure the impact of those AI-shaped journeys is a joke.
On Thursday, over a dozen media buyers on a Midtown Zoom call admitted they have no clue what happens to their ad dollars once they’re funneled into Google’s or Meta’s AI-black-box content feeds. You want to know who’s clicking, why, or what drove that conversion? Good luck—your dashboard is showing you last-click attribution in a world that’s all about suggestion engines and LLM-powered chaos.
Every agency deck this fall is stuffed with fancy AI keywords and “optimized journeys.” But when you press these so-called 10x agencies for receipts—actual tracking, real customer pathing, not just a pretty heatmap—they’re either hand-waving or blaming privacy updates. It’s all cargo cult nonsense. The measurement tools haven’t kept up, and half the big-name analytics platforms are still peddling 2022-era heuristics as if they mean anything for today’s TikTokified, Chat-GPT-mushed, discoverability-first customer experience.
Here’s the truth marketers don’t want to admit: If you’re pouring budget into AI-driven media and using legacy metrics to justify it, you’re not optimizing—you’re gambling. Worse, you’re letting AI vendors grade their own homework. If you care about actual ROI, you need to start building measurement in-house or demand real event-level data from your platforms. Otherwise, congrats: you’re the mark in the world’s biggest shell game.
Uncomfortable recommendation: Stop trusting agency decks and platform dashboards. Hire someone who can interrogate raw logs, instrument every touchpoint, and build your own damn attribution model. Yes, it’s harder than copy-pasting a Google Analytics tag. That’s why it works.
