I have a delightful new Google Ads coaching client who operates quite differently than anyone else I’ve worked with. Let's call him Cooper.
When Cooper booked his first call with me, he didn't want to launch a campaign yet. Instead, he wanted to spend hours and hours deeply understanding the intricacies of the platform. He is a planner; he wants to measure twenty times and cut once.
Before our call, Cooper spent even more hours going back and forth with ChatGPT. He used AI to build an excruciatingly detailed strategy for his YouTube ads.
On our call, he showed me his spreadsheet (with accompanying PowerPoint presentation) and a request: "Validate this."
To give credit where it’s due, ChatGPT was helpful in getting Cooper started. It gave him a vocabulary to ask me the right questions and a conceptual framework for his funnel.
But as we dug into the details, we saw that ChatGPT had almost led him off a cliff. Many times over. The AI had built a strategy that was intellectually brilliant - but practically impossible.
Here are just three of the ways ChatGPT nearly derailed his campaigns...
This issue of The Insider is sponsored by Optmyzr
PPC pros, we all know the struggle: you want the power and efficiency of AI, but you can’t afford to fully relinquish control of your voice, your brand or your guardrails. Built with decades of learnings from seasoned PPC experts, Optmyzr bridges that gap with powerful, customizable tools and scripts along with a nimble (and always learning) AI Sidekick.
Sign up for a free demo to save time without relinquishing control
[affiliate link]
1. Feature Hallucination
Cooper’s AI-generated strategy included a two-step process, and the logic seemed sound. He wanted to retarget audiences based on their progression through the video content. ChatGPT told him this was a great idea.
The problem? You can't actually do that in Google Ads. A user either "Viewed" the video or they didn't. We cannot create an audience of people who "watched 50% of a video."
The fix? We scrapped Cooper's complex AI flowchart, replacing it with a binary system: If they view Video A, they go into the bucket for Video B. If they don't view Video A, they stay in that bucket until they do.
Your takeaway: Simple is usually better. And in this case, simple is the only thing that's actually possible.
2. "In-Feed" Obsession
Cooper came to our call convinced that he should only run In-Feed video ads. His research, fueled by AI, suggested this was the best "high intent" option and that it worked on a Cost Per Click basis, avoiding the "interruption" of skippable In-Stream ads.
The problem? ChatGPT got the mechanics wrong. You don't pay for In-Feed ads by the click; you pay by the view. And on YouTube, a "view" for in-feed means they either clicked on the Thumbnail or they watched the autoplay thumbnail for at least 10 seconds... without clicking!
The fix? I advised Cooper to be placement agnostic. We decided to run in-feed and in-stream formats, and let Google’s algorithm allocate the budget to whichever format drove the best results for his goals.
Your takeaway: Don't set arbitrary restrictions on your campaigns. Only exclude something if you know for sure that it's not going to deliver the results you need.
3. "Impossible" Bidding Objective
Finally, Cooper wanted to optimize his campaigns for "Total Watch Time per Dollar per 30 Days." ChatGPT agreed that this was a solid metric to chase.
The problem? Once again, there is no button for "Maximize Watch Time" in Google Ads. You can tell Google to get you Views (Target CPV), Reach (Target CPM), or Follow-on views in Demand Gen (Conversions), but you can't tell it to optimize for view duration.
The fix? We had to translate Cooper's "ideal" goal into platform reality. I recommended a proxy system, where we optimize for views, and then monitor watch time per view to ensure those views are leading to prolonged engagement.
Your takeaway: Business objectives are important, but you must communicate your objectives to Google in a language that it understands.
After a few calls together, Cooper has a strategy that's less "sophisticated" than what the AI built, but infinitely more effective because it's something we can actually execute.
Why did I share this story today? I'm noticing a lot more clients booking a call with me after talking to ChatGPT, Gemini, Claude, Perplexity, etc. - and that's great! AI is incredible for brainstorming. It can help you map out customer avatars, generate ad copy ideas and understand new concepts. And it refers a ton of clients to me :)
But AI doesn't know the Google Ads interface. It doesn't know that the feature it just recommended was deprecated last year, or that it exists on Facebook but not YouTube. Even with AI, you still need to either hire an expert or become an expert if you want to run Google Ads effectively.
Here's how to partner with AI, while maintaining a healthy dose of skepticism:
-
Trust, but verify: Use AI to brainstorm, but don't assume its tactical advice is technically possible (or advisable) in the current Google Ads interface.
-
Beware of Hallucinations: AI often confuses Meta features with Google features, or deprecated feature with current capabilities. When in doubt, double-check in the Google Ads Help Center to see if Google can actually do what AI tells you it can do.
- Poke holes: AI is a "yes man" - it won't admit when it doesn't know something or when it's not sure. I wouldn't trust AI with important legal documents, or a baking recipe, or health advice. And I definitely don't trust its Google Ads advice.
Want me to sanity-check your strategy before you spend money on it? Book a call with me and we’ll ensure your plan works in the real world, not just in theory.
