One Extra Click

Like it or not, Google has a business to run and it is not going to let something as simple as the desires of its end users get in the way of all that.

Which is why, when Google announced plans this week to extend its Web search via labs, I naturally had concerns. Were they going to destroy &udm=14?

(Side note: Not the best timing for that—it was just featured in The Washington Post!)

I have been futzing around with the new view they just added, called “Web Guide,” and I will report that the good news is that it’s a fairly non-destructive implementation. They did not break the existing tooling to add this new feature. If you do a search with the &udm=14 modifier, especially using a secondary technique like your browser or udm14.com, it’s still going to work, at least at this time.

But the bad news about all this is that Google is still trying to add AI to their one small respite against AI. Let’s talk real quick about what Web Guide does. Essentially, I would describe it as a way to get the top results on a bunch of different topics at once. So, if you search for “salamanders,” it not only gives you the basic details for salamanders, it also does starting point searches for you for topics like “salamander biology and composition,” “salamander images,” “salamanders as pets,” “salamander mythology and folklore,” and even “salamander” in relation to Warhammer 40k. As you’ll see from my example below, Gemini actually pulled up the top two results of my search twice:

Compare that to the results I get for a standard search:

Essentially, the Web Guide is a bet that you’re going to want the top two results in a bunch of terms rather than the top ten results in a single one. It does not let you scroll to secondary pages. It is essentially a single view overview that tells you the absolute basics about five things rather than getting different results in a various number of topics.

If this was the version of AI search Google gave us last year, I think people would be a lot less upset with them. There would be fewer glue on pizza controversies, fewer complaints that Google was trying to discourage clicks. Sure, the publishers would still be mad as it would fundamentally change what a Google search represents. But it would still make sense to do it this way.

But instead, Google went the way they did with AI overviews, and it feels like this does not exactly engender confidence that the company will leave its AI-free search results unadulterated long-term.

Thinking about Google in terms of zombies

To sort of close the loop on this thought, I want to point out a piece by Emanuel Maiberg over at 404 Media. Emanuel edited my Motherboard pieces for many years, so I’ll flag that we’ve worked together a bunch. But I think his story, titled “Google’s AI Is Destroying Search, the Internet, and Your Brain,” is important here.

Based on a data point about how people are less likely to click on a link when they see an AI overview, the story helps to highlight how AI just ruins what made Google searches good. It’s not just what Google adds to them—it’s how it obfuscates original reporting and even favors LLM-generated content over the real thing:

Gaming search engine optimization in order to come up as the first result on Google regardless of merit has been a problem for as long as Google has been around. As the Pew research makes clear, AI Overview just ensures people will never click the link where the information they are looking for originates.

We reserve the right to whine about Google rewarding aggregation of our stories instead of sending the traffic to us, but the problem here is not what is happening to 404 Media, which we’ve built with the explicit goal of not living or dying by the whims of any internet platform we can’t control. The problem is that this is happening to every website on the internet, and if the people who actually produce the information that people are looking for are not getting traffic they will no longer be able to produce that information.

He goes further, pointing out that AI is highly exploitable, explaining how an artist invented a technical concept whole-cloth, only for the AI overview to digest it. This means, if you rely on Google’s AI alone to provide you information, the information is susceptible to manipulation or worse. While Google criticized the “flawed methodology and skewed queryset” of the Pew research in a comment appended to the piece, Emanuel’s point stands:

I wish I could say this is not a sustainable model for the internet, but honestly there’s no indication in Pew’s research that people understand how faulty the technology that powers Google’s AI Overview is, or how it is quietly devastating the entire human online information economy that they want and need, even if they don’t realize it.

When I posted Emanuel’s piece (I added it to udm14.com because of how good I thought it was), I came up with a metaphor that sort of nails down why stuff like &udm=14 is only a band-aid on the real problem: Just because you protected yourself from the zombies doesn’t mean the zombie problem itself doesn’t exist. The zombies are still wrecking society.

You can switch to Kagi or DuckDuckGo and any host of other search engines, but if people are using the AI results on Google without the right level of skepticism, they will be in danger of being harmed by them. It‘s not a sustainable situation, living life as if zombies do not exist. Society has to deal with the zombie problem, or it’s just going to fester and lead to gradual disaster.

As I’ve made clear, I am not an AI hater. But I consider Google an essential tool. And it bothers me deeply that they’ve just kept making it worse. Their implementation feels irresponsible.

The hype around AI is one thing. But the unwillingness to not give users choice in the matter, to implement destructive defaults, where people have to rely on hacks to fix them? That’s the real problem.