Let’s Start With “What Is the AI Overview?”
If you’ve searched for anything in the past year on Google that was more informational (think, how to do something or the meaning of a word), then you most likely have encountered AI Overview. There is a little symbol at the top of the search results that tells you that’s what it is:
But sometimes it’s easy to overlook this. So, basically, when you get a paragraph at the top of the search results that gives you an answer, most of the time a complete answer without you having to click a thing, that’s AI Overview. Google itself says that it works “by leveraging a large language model (LLM), like Google Gemini, to generate summaries of information from multiple sources on the web, including Google’s own index.”
The thinking is that, of course, Google already has all of this information, so what it is serving me in the AI Overview is just saving me a ton of time searching around the web. And that’s true. Google is giving you an answer, and you don’t have to click to go anywhere else. But how accurate is that information? In a lot of cases, that information is very accurate. Then, there are cases where it’s either outdated or not accurate at all, and far too many people are just accepting what is there and moving on.
The Wake-Up Call
Last year, I had a significant college reunion that was to take place in June. I was definitely planning to attend, but didn’t feel any rush to register, as there didn’t seem to be a deadline on the reunion website. There was a list of who all was attending on the site, and two weeks before the reunion, I got a call from a close friend telling me that the registration deadline was earlier that day and that I had missed it.
Feeling like I would have to beg and plead, and ready to do so, I called the Alumni Association the next day to see if they would let me register post-deadline. I was told that I could actually register up until the very first day of reunion on the website. Relieved, I called my friend, who then told me that he had Googled it, and that is what AI Overview said — that the deadline to register was that day. Turns out, that was the last day to request a refund, which was mentioned on the site, but not the last day to register. AI Overview got it wrong.
While this incident wasn’t life-altering, it made me think that some could be. Granted, AI Overview had just launched, and some inaccuracies were probably to be expected, but it taught me that this new piece of technology needed to be fact-checked from the get-go.
Google Should Know Its Own Stuff, Right?
Being in the paid search field for as long as I have (10+ years), I’ve seen Google Ads go through so many changes, updates, revisions, designs, and ideologies. It’s impossible to know every given thing about every given aspect every day, as there were probably changes yesterday. So, when a new feature shows up in the Google Ads platform that wasn’t there before and had no “announcement” that it would be showing up, what better way to figure out how to use this new feature than to Google it? You figure that’s a Google product, so Google AI Overview should have the answer, or at least the link given at the end of the answer should lead us to the source of truth, right?
What many of my colleagues and I have found is that AI Overview’s answers to questions related to Google Ads are consistently either inaccurate or outdated, and the link given at the end of the answer sometimes leads to a page that not only doesn’t match the AI Overview, it contains information that has nothing to do with the original question. This happens time and time again, but here is just the latest example:
A client recently asked about negative keywords for PMax campaigns. Negative keywords are used so that your ads won’t show when those keywords are searched. Up until recently, you could not have negative keywords at the campaign level for PMax, only at the account level, meaning it would affect all of your Google Ads campaigns. I knew that Google started rolling out this feature recently and wanted to see when it launched. So, I searched “pmax adding negative keywords.” Here is the AI Overview that served (this was two days ago as of this writing):
This is outdated information, as you can definitely add negative keywords at the campaign level for PMax now. (Side note: I Googled the same thing just now and got an accurate answer from AI Overview. So, it is learning.) If someone not in the industry is trying to set up their own PMax campaign and doesn’t have the experience of my reunion situation, they may set up a campaign that will cost them money where they don’t want it to be spent, simply because AI Overview was outdated.
I Don’t Hate AI Overview
After reading this, you may think I hate this feature. I truly don’t. It can be of great use and value if people who see it dig a bit deeper to make sure it’s accurate. And, as mentioned above, it is learning. So, answers will become more relevant as time goes on, but right now, AI Overview answers that can have an impact on your life or your wallet should not be taken as gospel. Even Google “sort of” admits this if you hit the “Show More” button at the bottom of the answer. You will/should see “Generative AI is experimental.”
Credit Where Credit Is Due
I Googled “take something as gospel.” Below is what I got. This one is spot on.
P.S. — This article was whipped up the old-fashioned way — by humans. Except where noted as an AI Overview response, no bots were used in the making.
Angie Batten
Latest posts by Angie Batten (see all)
- AI Overview: Don’t Take It As Gospel - May 12, 2025
- Google Ads: Your Power Tool to Promote Your Black Friday Deals - November 26, 2024
- PPC Updates: Confidential Matching to Protect User Privacy - October 22, 2024
Leave a Reply