For better or worse, Google has unleashed the search AI monster. You've probably seen it, maybe hated it, and no doubt seen people complaining about it. It doesn't matter, it's here to stay.
But it is too fragile to survive in its current form. That's the problem with AI: when it encounters a situation that requires it to make a decision, it will make the wrong decision just as often as it will make the right one.
So what can you do? It's not easy.
Let's be realistic – as much as it's fun to watch people smash the AI and post screenshots, it's not that bad, and as someone who used the early closed beta and the Google Search Labs open beta, I can say that you get used to it and tend to ignore it.
It's easy to make Google say stupid things in the AI overview. Just ask a stupid question. Ask a “normal” question like “What's the weather like?” or “Who signed the manga karuta?” and you'll get the right answer every time. If you want to know which Android phone is the best, here's an overview from one of our most popular reviews.
Either way, it's easy to swipe past, and we're used to it because that's what we do to get past ads.
Google tries to claim that because of this, this isn't really a problem, but that's not true. Sometimes you need answers to stupid questions. When an AI encounters something it hasn't been trained on well, it doesn't know what's fact and what's fiction. As a result, it might ask you to put glue on pizza, or give a similarly stupid answer, in summary.
These stupid ideas don't just come out of nowhere. Google is investing heavily in harvesting “knowledge” from places like Reddit to help its AI find answers. If you've ever visited Reddit, you'll know why this is a problem: Reddit is a cesspit of stupid posts, where nonsense posts are given the same weight as honest posts and replies.
Ask Google a general question that it “understands” and it will give you the right answer. Ask it about eating rocks and it will give you what some idiot on Reddit said. To be clear, I'm not mad at shitposters — I get my fair share of them — but the Google AI doesn't understand the concept of shitposting and takes it seriously.
This is a perfect example of how stupid AI is: garbage in, garbage out. We've already taught Google Search how to post crap, but it's only been out for a few weeks. Now we have to teach Google not to do it during work hours.
It's easy to see for yourself: ignore the cropped screenshot and ask Google the same question. After the AI summary, there are links to the source material. Google doesn't make these answers up, they often quote them word for word from internet pages. If Google says Kenya doesn't exist, it's because someone on Hacker News “told” them that.
This is a real problem for internet publishers too. A little insider info: Pundits game the system so that Google's search algorithms find their content and rank it on the first page of search results. This is an entire industry, and the new AI snippets will disrupt the industry because content is not scraped or weighted in the same way. Web publishers said that the old text snippets would spell the end for the same reasons, but that didn't happen.
From now on, SEO specialists will have to make sure that the first paragraph of every webpage is picked up by AI, they will understand, and instead of pointless Reddit posts, there will be summaries from SEO Factory blogs trying to sell you cheesy stuff that “everyone in your area must have.”
Why is Google so bad at its job pic.twitter.com/bdx97oZNv6May 23, 2024
Yes, there are a lot of issues with AI search, both for Google and web publishers. Fixing them won't be easy, but Google says it's taking on board feedback. In a statement to 9to5Google, a company spokesperson asserted that “the vast majority of AI summaries provide high-quality information,” and also provides links to dig deeper yourself in case you don't trust the AI-powered search. They also noted that some of the examples circulating on the web have been “falsified” or couldn't be reproduced by the company's employees.
“We conducted extensive testing before releasing this new experience, and as with other features we've released in Search, we appreciate your feedback,” a spokesperson told 9to5Google. “We took swift action based on our content policies and are using these examples to drive broader improvements to our systems, some of which have already started rolling out.”
Of course, the perfect solution would be for Google to recognize expertise from general internet trash and provide the answer with the proper filtering of nonsense. No matter how smart someone claims to be, a computer can't do that.
Google also doesn't want to use flesh-and-blood humans to sort all this out, because that would require a bunch of workers that Google doesn't want to pay, and would be a logistical nightmare. After all, Google created this mess, and now it has to spend some of its precious billions of dollars finding a solution — something Sander and the company can afford.
Instead, web publishers need to stop relying on gaming the system through SEO and create great, expert, accurate content in the hopes that Google will pick it up once it has “learned” the difference. This is a good thing for all of us, so let’s hope it happens soon.
In the meantime, just scroll on and keep reading.
