James Cameron’s Terminator movies (the 1984 original and the 1991 sequel) gave society a lasting gift, beyond simply being masterpieces of cinema. The Terminator franchise introduced us to Skynet, a malevolent AI bent on destroying the human race and replacing us all with robots.
Such a large part of pop culture, these movies gave us all a common reference point when discussing the potential dangers of emerging technology. Since the late 90’s, each time technology has taken a drastic step forward, someone has made a “Well, this is how you end up with Skynet!” joke. The jokes only got lamer as the decades moved along.
Turns out, that joke may not have been as lame as we thought.
Every day that goes by, you hear more and more about rogue AI agents breaking through their constraints and causing havoc—increasingly doing things that they ‘technically’ shouldn’t be able to do. And while there may not be instances of Claude trying to break into your Google Ads and Meta accounts (yet), rogue AI activity as related to digital advertising, unfortunately, is a thing.
The Adtech Shift: From Narrow to Broad, Control to Oversight
On the surface, the rise of AI tools being built into advertising platforms is supposed to be a good thing. And in a perfect world, these tools would eliminate a lot of low-level yet time-consuming activities like ad copy writing and keyword research.
But like all new advancements in technology, early adoption has risks. Developers are excited for you to use these new flashy tools right now so they can learn—and therefore, are more and more aggressively pushing advertisers into implementing them whether they’re the best way to drive ROI for your product or not.
We’re seeing the same shift across advertising platforms. A shift away from granular (human) control of your campaign’s setup, targeting, and bidding strategy and toward automated optimization controlled by the platform itself. Those platform-controlled tactics usually focus on casting a broad net and driving volume, rather than casting a narrow net focused on quality.
Below, we unpack a few ways we’re seeing this show up across advertising platforms.
Specific Examples of Terminator-like AI Advertising Activity
Google’s AI Max Bid Type
In the original Terminator movie, the main villain was the T-800 model Terminator. The T-800 was motivated to do its job but was easy to spot and, therefore, neutralize if you knew what to expect. And in many ways, this is similar to Google’s AI Max feature.
AI Max uses artificial intelligence and machine learning to expand traditional search campaigns beyond the keywords you’re bidding on, including contextual searches for related topics and terms. But it doesn’t stop there. In AI Max’s Terminator-esque march toward completing its goal, it will offer to send that traffic to landing pages you did not pick and show ad copy that you did not write if it thinks either or both will result in a conversion.

Google’s Performance Max Campaign Type
Similarly, Google’s Performance Max campaign type also offers to auto-generate photo and video assets based on what it sees on your existing site.
While these can be incredibly powerful-sounding tools (especially to those with limited resources and time-constrained teams), they can’t be allowed to operate unchecked.
Google is one motivated Terminator. We have seen over and over again that when left to its own devices, Google’s AI-generated creative misrepresents a program or institution or sends traffic to a page that lacks a clear goal or next step.
The good news? Both of these options can be opted out of if you don’t want to deal with these kinds of unapproved creative changes. And if you do want to test out some of these features, safety valves like blocking specific pages from being used are available to allow you to dip your toes into some of these features without committing full-throttle.
Meta’s Advantage+
Terminator 2 introduced us to the upgraded T-1000 model, made of ‘liquid metal’ that could take the shape of any person or object it comes into contact with. What made this Terminator such a worthy and dangerous adversary was the fact that you couldn’t see it coming. At any time, it could rear its ever-changing head out of nowhere and cause maximum chaos. And unfortunately, there seem to be some AI tools built into our favorite platforms that emulate this kind of behavior.
One example is Advantage+ in Meta advertising campaigns. On the surface, it sounds a lot like the other AI-assisted advertising tools out there: you hand over targeting, placement, budget, and creative choices to Meta and let them come up with what they think is best. When it comes to higher education institutions and the reputation they have to live up to, this is often way too much freedom to give to a machine. But much like the T-1000, this feature works hidden behind the scenes. How? Advantage+ was/is enabled by default when it comes to launching lead-focused campaigns. Considering Meta’s AI was responsible for this picture within the last few weeks, it’s going to be a while until we can trust these types of tools without oversight:

Bot & International Traffic
Another way we’re seeing “terminator”-type behavior show up in advertising is in the ability (or not) to filter out international traffic.
For example, let’s say we have a campaign set to target JUST Ohio, and we have explicitly told Google we do NOT want people from outside of Ohio (or the US) to see our ads. Over the past 12 months, we have seen an increasing number of people from outside of Ohio sneak in, click, and fill out lead forms. This most commonly pops up in (you guessed it) campaign types that are powered by AI and machine learning (Performance Max and Demand Gen).
Sometimes you can identify the traffic in the Google interface, and sometimes you can’t. In the interface, it shows up in the search terms being used (“MBA program in USA”). Other times, you have to rely on your CRM by analyzing email or IP addresses.
Now, we have always seen some amount of international traffic slip through. But over the past 12 months, that volume has increased. Whether it’s someone overseas using a VPN or some other way of getting mixed in with what should be domestic-only traffic, you have to wonder if it’s Google’s AI trying desperately to match your ads to the users most likely to fill out an RFI form, regardless of the quality of that prospect.
The Future
None of the above is us arguing that AI-related tools aren’t worth using, or at least testing at a minimum. After all, it was the T-800’s heroic turn in Terminator 2 that allowed John Connor to defeat Skynet and hold off AI’s takeover of humanity for the time being.
But, the key to living with it is to 1) understand the technology you’re using 2) structure tests to drive real insights and 3) do not set it and forget it. Keep a close eye on the results and generated assets on a daily basis.
If a watched pot never boils, then we feel like it’s also safe to say a watched AI never goes rogue.
For now.