The platform's AI works for the platform
Open Google Ads and the platform greets you with recommendations, each promising a better optimization score. Open Meta Ads Manager and Advantage+ offers to take the wheel. The advice is free, it is instant, and it comes from the company with the most data about its own auction.
It also comes from the company that gets paid when you take it.
The structural problem, stated plainly
This is not an accusation of bad faith. It is a description of incentives, and incentives are structural.
Google's recommendation engine is built, trained, and measured inside a business whose revenue is your spend on Google. Meta's is the same, for Meta. When a platform's AI weighs "suggest the advertiser raise budgets" against "suggest the advertiser move money elsewhere," the second option is not in its vocabulary. No platform recommendation has ever been, or will ever be, "your money would work harder on a competitor's inventory."
Each platform's advice is also structurally blind. Google cannot see your Meta results. Meta cannot see what your organic search already wins for free. Neither can see your actual revenue the way your analytics does. So even a well-intentioned platform recommendation optimizes a slice of your marketing while being unable, by design, to consider the whole.
What platform recommendations are genuinely good at
Honesty cuts both ways, so here is the other side. Platform recommendations are useful for platform mechanics: fixing disapproved ads, filling thin ad groups, adopting formats the auction genuinely rewards, repairing broken tracking on that platform. The platform knows its own machinery better than anyone, and advice about the machinery is worth hearing.
The pattern to watch is the category of advice: mechanics versus money. Recommendations about mechanics are usually worth a look. Recommendations about money, raise this budget, broaden this audience, remove these constraints, expand to these placements, are the ones where the advisor's paycheck and your interests pull in different directions.
A practical rule many experienced buyers use: never auto-apply, read everything, adopt mechanics freely, and treat every money recommendation as a sales suggestion that must re-earn its case from your own cross-channel numbers.
What neutral looks like
A neutral layer has three properties no platform can offer:
- It sees everything. All channels, plus your analytics and search presence, in one analysis, so "where would this money do more" is a question it can actually ask.
- It sells nothing. No inventory, no auction, no revenue from your spend. Its only way to win is to be right.
- It shows its work. Every recommendation carries the evidence and the money at stake, so you can disagree with it from the same numbers it used.
The platforms' AI is not going away, and you should keep reading it, the way you read any vendor's proposal: with interest, and with your own numbers open in the next tab. The job of a neutral decision layer is to be that next tab, permanently.
OmniCommand reads every channel, sells no ad inventory, and attaches the evidence to every recommendation. Early access is open at omnicommand.ai.