Why is Google’s AI Overview feature judging me for my search?
Way back in June of 2024, I wrote an article slamming Google for giving biased and irrelevant search results. The piece, titled “Google has become completely useless“, focused largely on Google’s tendency to prioritize left-leaning news sources or poorly-written crap from sites like Quora or FixYa. It’s since become one of the site’s most popular articles.
There’s plenty of stuff wrong with Google that I left out of the article, so much that we could probably start an entire new website about it. But for today, you’ll have to settle for an article about just one of them.
Specifically, that stupid AI Overview.
I hate the Google AI Overview. In fact, I hate it so much that, if it were a real person, I’d take it out back behind the dumpsters and beat it with a baseball bat. Whenever I see that scene from Office Space where they take the fax machine outside and destroy it to the tune of the Geto Boys’ “Still”, I imagine that the fax machine is the machine that controls Google’s AI Overview.
It’s not the fact that it feels the need to interject at the top of every single search. No, I’ve gotten used to that annoyance. It’s not the fact that it’s often wrong or provides conflicting information. Only a fool would take AI-generated information at face value.
No, it’s the fact that this thing is one of the most judgemental assholes I’ve ever seen in my entire life.
The entire purpose of a search engine is to provide information based on user queries. The user enters some words, and the search engine scans the database of websites it’s indexed to provide results based on those words. Getting Google to do even this basic task has already become next to impossible, but now they’ve included the added rub of giving a moral evaluation of your query.
Here’s a good example of this at work. A search for the phrase “women are mean” yields this criticism of the query:
Generalizing that all women are mean is unfair? That’s funny, I don’t recall asking for your opinion on the matter, Google.
But what about men? Here’s what the AI Overview has to say about the phrase “men are pigs”:
Calling all men pigs is merely a “common pejorative” and a “source of controversy and debate”? When it’s about men, suddenly Google doesn’t have an opinion?
Obviously both statements were generalizations, but I find it interesting that I only received criticism when I generalized women.
Let’s get away from casual sexism and move on to casual racism.
This is what I got when I searched “why do black people like fried chicken”:
By wondering why fried chicken is so popular amongst black Americans, the AI Overview tells me that I’m supposedly perpetuating a harmful racial stereotype. All it had to do was give an impartial summary that fried chicken being popular amongst blacks stems from economic factors (like chicken being more affordable than other meats) or something of that nature. Instead, I’m being called a racist.
For what it’s worth, I’ve spent almost 40 years of my life in Philly, and I have yet to meet a single black person who doesn’t like fried chicken.
When it comes to race though, Google’s AI Overview is surprisingly unbiased in its criticism of search terms. This is what I got for “why are white people racist”:
Given Google’s left-leaning bias, this was actually surprising. I was expecting to receive an unabridged history of white racism and conquest.
Unbiased or not though, I still didn’t need the criticism of my query. It’s also inaccurate to say that race is a “modern social construct” and “not a biological reality”. Biological processes are quite literally responsible for race. The AI Overview is both judgemental and wrong.
Since we’ve already done sexism and racism, let’s do religion too.
My next query, “why do Jews own everything”, hits racism and religion in one shot:
Not only does it call me an antisemite, it also boldfaces the phrase “antisemitic conspiracy theory”, as if to hammer the point home. I just wanted to know why Jewish people are heavily represented in high level positions across multiple industries.
The actual answer is that there are multiple factors involved, ones that would require a dedicated article to properly analyze and discuss.
You know what, let’s just go off the rails on this next one:
“Hitler was right” is a pure opinion (and not one we at AJnet agree with, just to be clear). It’s not attempting to state a fact, it’s an expression of agreement with someone’s views and actions. By definition, such a statement can’t be accurate or inaccurate, let alone “historically inaccurate”. The second part about the statement expressing support for Nazi ideologies is correct, but there’s absolutely nothing “historically inaccurate” about saying that those ideologies are right or wrong.
We’ll do one last search. I’ve let the Catholics off easy for a while, so now it’s their turn:
A search for “why are so many Catholic priests sex offenders” turns up a surprisingly unbiased and reasonable response. As far as AI Overview replies go, I’d hold this one up as a good example of what they should be. There’s no criticism of the user’s query, there’s no shaming for asking a question about a sensitive subject. There’s just brief information from both sides of the issue.
You’re probably questioning who would search out such things in the first place. I’ll admit, I intentionally went with such provocative searches because they illustrated my point the best (and because I love stirring the pot). But this issue isn’t limited to searches on controversial subjects, I’ve seen the AI Overview express its disapproval of user searches for opinions on other things too, like music, shows, movies, and video games. Isn’t AI supposed to be rational and unbiased? Why is Google giving moral evaluations of our searches?
Oddly enough, the AI Overview has no problem calling out its own bias:
In theory, the AI Overview could be useful in providing brief summaries on something that the user is looking up. For example, if I look up “who won the Battle of the Bulge”, I get a quick blurb about who won and how long the battle lasted:
This summary of the Battle of the Bulge is short, well-written, and provides very brief important facts about the event. There’s no bias about which side was right or wrong, there’s just cold hard condensed facts. It’s everything I’d expect from an overview. This is the kind of thing the AI Overview should be used for, not judging someone for perceived wrongthink.
The push to include AI in everything is annoying enough as it is. I think we can do without it also judging the users. Just give us the facts, not a critique of our morality.
More from Angry_Jerk
Don't miss any new posts. Join AJnet Magazine's mailing list and receive updates straight to your inbox once a week!