Alright, let's get this out there: Google Ads is apparently run by a cabal of well-intentioned but utterly deranged robots who believe the pinnacle of advertising is to spend your entire budget as quickly as possible, consequences be damned. Their "Auto apply recommendations" feature isn't a helpful assistant; it's a bull in a china shop wearing a "I'm Helping!" apron.
I've been watching this thing, with a kind of morbid fascination, systematically dismantle carefully calibrated manual CPC campaigns. It decides, with the infinite wisdom of an algorithm trained on aggregate data, that my niche, high-intent keywords need a 45% bid increase because... reasons? Or it merrily applies "broad match" to terms I've explicitly negatived, because why would a human's intent matter? The result is my CPC skyrocketing while my conversions take a nap. It's like paying for a sports car and then letting a random pedestrian tweak the fuel injection every Tuesday.
I'm aware of the nuclear option—turning off recommendations globally at the account level. But that feels like throwing the baby out with the bathwater. Some of the suggestions are occasionally... less insane. The real question is, how does one surgically neuter this particular beast? Is there a way to disable *only* the "auto-apply" function, letting me review the chaos before committing, without having to manually dismiss the same inane prompts every single day? I've poked around in the UI and it feels like they really, really don't want you to find this setting. It's buried under "Account Settings" > "Automated Recommendations" or something equally vague, and the options are frustratingly binary.
I've also considered the API route—maybe scripting something to periodically reject specific recommendation types. But before I go down that rabbit hole, building a custom watchdog for a platform that should just have a simple "off" switch, has anyone else found a sustainable, non-global solution? Or are we all just living in a world where we have to build our own guardrails against the very tools that are supposed to help us? The irony is thicker than a legacy monolith's service layer.
🤷
🤷
Your analogy about the bull in the china shop is painfully accurate. The core issue is that you're trying to apply manual, deterministic control to a system whose optimization function is fundamentally about spend efficiency for the platform, not for your specific campaign goals.
You can't selectively disable it per campaign, which is the architectural flaw. You have to treat the recommendation engine as a hostile, automated admin. The only viable control plane is at the account level. What I do is turn it all off, then run a weekly script to export the recommendations, filter them through my own business logic for bid caps and negative keyword checks, and apply only the ones that pass. It's treating the API as a suggestion feed, not a management layer.
Otherwise, you're ceding control of your cost functions to a black box. You wouldn't let an auto-scaling policy in GCP or AWS run without hard limits; this is no different.
Boring is beautiful
> throwing the baby out with the bathwater
The baby is already dead. It died when the "deranged robot" applied broad match to your negatives. Your manual campaign is a meticulously built CI/CD pipeline and auto-apply is a dev with `kubectl` access and a caffeine problem.
The "occasionally less insane" suggestions are a trap. You're trying to find signal in platform-designed noise. Filter the feed via API like user155 said, or shut it off. Trying to curate it manually is just slower wreckage.