AI Can Get Your Google Ads Campaign Close. Here’s Why That’s Dangerous.
A client found me a few weeks ago by typing exactly what she needed into AI, in detail, until it pointed her toward hiring someone like me. There's something funny about that. She used AI well enough to know she needed a human.
I was so excited that I called all my PPC industry friends because my content and my writing is attracting EXACTLY my ideal client. I am proof that AI works but there is more that was revealed when this client found me.
What had her stuck wasn't a huge structural problem in her Google Ads.
It was specific recomendation. AI had told her a form fill isn't a good conversion, and she didn't know whether to believe it.
She had the self-awareness to feel unsure instead of just changing it and optimizing her Google Ads for a new metric, which turned out to be the right instinct. That advice about Google Ads from AI isn't wrong. It's also not right.
It depends entirely on her sales cycle, her funnel, and what a lead is actually worth to her business — context AI didn't have and never asked for.
That's the real danger here, and I want to be clear about what I'm not saying. I'm not anti-AI.
I use it constantly, for macros, for blog posts, for decks, for explaining tactical concepts faster than I could type them myself.
The danger isn't that AI gives bad advice. It's that AI gives confident advice, and confidence is the one thing that should never get mistaken for accuracy when real marketing budget is on the line.
Where AI Advice Goes Wrong In Google Ads
This is something you might not notice so this isn’t meant to make anyone feel bad or to tell you to refrain from using AI.
Again, I use AI all the time to help explain tactical concepts, and if you use it enough you start to notice the same defaults no matter what you ask: add negative keywords, add long-tail keywords.
That used to be close to universal advice.
It isn't anymore.
Most long-tail terms now carry so little search volume that building a strategy around them barely moves anything in a modern Google Ads account.
The real answer today lives somewhere between Performance Max, the newer AI Max, and how much risk you're willing to take on with keyword strategy itself — this a far more nuanced call than "add more long-tail."
AI keeps handing out yesterday's best practice because that's what it was trained on, not because it's trying to mislead you.
I saw the same staleness when I built a fake advertiser and asked Gemini to construct the account from scratch and tell me exactly which campaigns to choose.
If you want to watch my YouTube video I covered it here in my Google Ads Unfiltered episode:
The Structural Red Flag Of A Google Ads AI Recommendations
I used AI to test what it would recommend in a fictional scenario. It recommended Standard Shopping campaigns with siloed, generic Search ad groups — a setup straight out of 2018.
That structure starves a new account of the broad signals modern Smart Bidding needs to learn. Ignoring Performance Max in favor of Standard Shopping isn't a stylistic choice anymore. It's could be a strategic failure dressed up as caution.
But other days, I need a Standard Shopping campaign. In fact I have it running in a campaign I am consulting on currently.
My own setup however runs the opposite direction: secure brand terms first, then launch PMax with strong assets and a clean product feed so the system can learn fast. Only once that core conversion data is solid do we expand into non-brand Search, because PMax already leans hard into non-brand, YouTube, Discovery, and Display on its own.
But this is my style and it is very case by case.
The Bidding Red Flag With AI Recommendations
When I gave it no conversion data, it suggested manual bidding with a $5 max CPC, with no way of knowing whether $5 was even close to correct.
Manual bidding tells Google "just get me clicks," and it leans on a last-click model that ignores the rest of the customer journey. For a business chasing profititablity instead of traffic, that's a real cost, not a quirk.
What I do instead: start on Maximize Clicks just long enough to generate data, then move to a value- or conversion-based strategy — Target ROAS for e-commerce, Target CPA for lead gen — once there's enough signal to trust. The target itself comes from your margins, not a guess and not from what AI says.
Bonus: How to Set Target ROAS and Target CPA
I wanted to present information for how I do things in a real account below. This is the thinking I go though in terms of setting targets in my ad account,
| Strategy | Goal | Formula | Example |
|---|---|---|---|
| Target CPA Cost Per Acquisition |
Maximize the volume of conversions at a specific cost. | Avg. Conversion Value − Profit Margin − Shipping/COGS = Max CPA | Avg. price $200, COGS $70, desired profit per sale $80 → Max CPA target = $50 |
| Target ROAS Return On Ad Spend |
Maximize the value of conversions at a specific return. | (1 / Max Ad Spend %) x 100 = Target ROAS % | Max ad spend 25% (willing to spend $1 in ads for every $4 in revenue) → Target ROAS = 400% ($4 revenue for every $1 spent). |
The AI Ads Advice That Sounds Accurate And Is Not
The trickiest version of this isn't when AI is flatly wrong. It's when AI is surface-level right.
I had a client whose non-brand campaign was running a cost-per-acquisition north of $300, while a competitor campaign converting non-brand traffic for under $10.
Ask AI why, and you'll get a plausible answer: the account is still in "exploration mode" and hasn't found the right signals yet.
That's not materially false.
It's also not the real answer. It’s like saying “my kid has not grown up yet” which is why they are misbehaving.
The real reason is that, in a small, heavily competitive local market like this one, people search for the competitor by name far more than they search the generic term, because the category isn't written about anywhere online and there's no broad public demand to anchor a non-brand strategy to.
AI can't know that. It isn't a market with enough data exhaust for any model to have learned it.
That's the actual risk: not that AI is wrong, but that its wrong answers and its right answers sound identical.
The Targeting Red Flag
AI also told me to use Custom Audiences inside Search campaign targeting.
Custom Audiences are built for Display, YouTube, and Demand Gen, not Search, which runs on intent (what someone types) layered with audience signals, not interest-based targeting.
Suggesting Custom Audiences for Search means the model is blending best practices across totally different platforms — technically unworkable, and a real way to waste a setup window you don't get back.
It also generated ad copy I'd never run for a premium brand without a second look — phrasing like "Used by Michelin Star Chefs," a claim that's one unhappy chef away from a legal problem if it isn't airtight.
AI can write fast. It can't vet a claim for whether it's actually true, or whether your legal team would let it through.
Where This Leaves You
Here are my thoughts I don't think there's a great answer here: we're all using AI more than we probably should trust it, and we're all figuring out in real time how much to lean on it versus when to override it. That's not a you problem. It's the moment everyone running a business right now is in.
What I do know is this: AI will get you to average. It's trained on what most businesses, in most markets, mostly do. Your business isn't average. Your margins aren't average, your market isn't average, and the gap between "close" and "right" is exactly what a $300 CPA versus a $10 CPA looks like in real dollars.
If you've built a strategy with AI, or inherited one from someone who did, a Protective PPC™ Assessment will tell you which parts are solid and which parts are average advice wearing a confident tone.
That's exactly what the $750 Google Ads Audit is for: a deep review and a recorded walkthrough so you know exactly where your strategy, or your AI's strategy, is falling short of the modern standard.
