For almost fifteen years, you could be a pretty mediocre marketer and still get by. Meta and Google were so dominant that "channel strategy" basically meant deciding what percentage of your budget went to each. Brands optimized within walled gardens, ran the same playbook your competitors ran, and as long as basic best practices were followed, the platforms did most of the work.

That era is over, and a lot of brands haven't caught up yet.

This week, reports confirmed that Meta is going to pass Google in total digital ad revenue in 2026. $243B vs. $239B. First time it's ever happened. Meanwhile, OpenAI is quietly rolling out an ads manager, dropped its pilot minimum from $250K to $50K, and hit nine figures of annualized revenue six weeks into testing. Search itself is fracturing, now happening inside ChatGPT, Perplexity, Reddit, and TikTok. The "search box" is everywhere now, and Google doesn't own most of those boxes.

The story isn't really Meta beating Google; it's that the entire ad ecosystem is reorganizing in real time, and the reflexes most teams built during the duopoly era are the wrong reflexes for what comes next.

What the duopoly trained out of marketers

When two platforms own 60%+ of digital ad spend, the optimization game is internal. Instead of testing new channels and formats, brands made tweaks to allocation between Facebook and Google. Testing became something so many brands did to creative variants and audience segments, not to entirely new platforms.

That's not a moral failing. It's what the incentive structure rewarded. Why pilot a $50K test on some unproven platform when you can put the same money into Advantage+ and get a predictable return tomorrow?

But underneath, most marketing orgs lost the ability to test net-new channels at speed. The processes atrophied as the internal appetite for unproven spend dried up. So when a real channel shift starts happening (like a serious chunk of high-intent commercial search migrating to ChatGPT and Perplexity), the org doesn't have the existing practice to move on it.

We've been having this fight for years

This isn't a new conversation. The same dynamic played out with CTV, retail media, and TikTok. Clients who moved early built advantages that are still paying out. The ones who waited showed up to a more crowded auction at a higher price.

It's happening again right now with ChatGPT ads. The clients who are growing carved out 5-10% of media spend for emerging channels and got into a pilot last fall. They moved early, locked in favorable unit economics, and are already building learnings.

The clients who are stagnating said, 'Let us know when it's proven.' By the time it's proven, the CPMs have tripled, and the window is gone.

The data backs it

Keen Decision Systems just published an analysis of 455 brand models covering $42B in spend. Brands they classified as "winners", 5%+ growth in both sales and NPV over 12 months, expanded their channel mix at twice the rate of stagnant brands. They didn't spread thin. They anchored on what worked, then layered new channels in once thresholds were hit.

The inverse pattern is more telling: 64% of marketers justified new tactics with publisher case studies after the fact. Two-thirds of the industry isn't actually testing; they're rationalizing. They wait for a competitor to validate the channel, then pile in late at peak prices.

What to actually do about it

The old discipline question was: "How do I squeeze more out of Meta and Google?" The new one is: "What's my rate of valid experiments per quarter, and is it accelerating or decelerating?"

A few things worth doing this quarter:

Audit how much of your budget is in "test and learn" vs. "scale and optimize." If it's under 10%, you're under-investing in optionality. If your team can't tell you the number, that's the bigger problem.

Look at your last four quarters of channel mix. If it's basically the same as it was in 2023, you're not stable, you're stuck. The underlying ecosystem has moved.

Talk to your finance partner about what a "valid loss" looks like on a channel test. Most CFOs have never had this conversation. They evaluate test spend the same way they evaluate scaled spend, which is a category error that kills experimentation.

If you haven't gotten into ChatGPT ads yet, now is the time. The pilot threshold dropping to $50K isn't an accident; OpenAI wants advertisers building learnings before the floodgates open. We've had early access since the pilot launched and have been running campaigns across several verticals. The early results are worth a conversation. The cost of being early is going to look trivial compared to the cost of being late.

Jackson Richards, VP of Strategy