Another twist. Bigger this time.

Google is changing things, and it's creating trouble for newspapers. Don't use AI to verify images or facts from the riots in Los Angeles, and Wikipedia has gotten an AI scolding from dedicated volunteers.

This is Kludder!

It's been more than a quarter of a century since the first Google search took place. Google founder Larry Page had the honor, on July 8th 1998, and with that, the internet became accessible for everyone to surf on. Today, Google is searched 9.5 million times per minute.

While Google grew rapidly, it became absolutely crucial for websites to be visible online. Everything was about landing that top spot on the search results. If you were listed on Google's second page, you might as well not bother.

Then ChatGPT burst onto the scene.

From search engine to answer engine

Google is feeling the pressure. After years of monopoly in the search engine market, something is happening. And it's all thanks to artificial intelligence. For the first time, the tech-giant is experiencing a decline in the number of searches.

ChatGPT has been connected to the internet for a while now, and so we've been able to get fresh information, right in the chat - naturally sprinkled with some hallucinations and falsehoods. In addition, companies like Perplexity have taken up the fight with Google, and their goal is to retire googling altogether:

Google offers what we can call a passive search on the web. You type in a question, or keywords, and Google delivers the links it thinks - based on an insane amount of data about you - will be relevant. But you have to quality-check and go through these links yourself.

Perplexity, and eventually language models like Claude and ChatGPT, do what they call an active search. By reading through and compiling information, the AI tool can give you a short summary with references to various sources it has used.

Google's desperation

To keep up with this development, Google released a service last year called AI Overview. You may have seen it, a small blue star next to a brief summary of exactly what you were wondering about.

But AI Overview didn't work quite as it should. Among other things, by suggesting using glue on pizza so the cheese stays in place. Since then, Google's AI Overview has been somewhat troublesome, and there has been a lot of wrinkles to iron out. In addition to that, the services from Perplexity and the language models simply seem more modern and engaging to use.

That's why Google launched a new and improved search experience, with the name AI-mode. The service doesn't exist in Europe or Norway yet, and was launched in the US as recently as May. AI-mode connects Google's own language model, Gemini, and the search engine. In this way, you should get a two-way conversation and take googling to a new level.

The whole thing takes place in a chat window where Google can suggest relevant links that you can click on or ask follow-up questions about. And this is where the story might've ended; Google keeps it hegemony and continues into eternity.

But while AI Mode might give Google a brief respite, it's turning the lives of other companies upside down.

The media houses' Google collapse

Mail Online, the digital venture of British The Daily Mail, has seen a brutal drop in the number of visitors to their pages from Google. Director of SEO and editorial e-commerce at Daily Mail, Carly Steven, explains the situation:

When we rank as number one in organic search, the click-through rate is about thirteen percent on desktop and around 20 percent on mobile. When we still rank as number one organically, but AI Overview appears, the click rate falls to under five percent on desktop and seven percent on mobile - Carly Steven to Press Gazette

The same article shows that AI Overview reduces clicks by 34.5 percent on average. And remember; Google's new AI Mode isn't even available in Europe yet.

Similarweb has conducted studies among selected American online newspaper publishers, and the numbers are dramatic. Business Insider, which has had to lay off more than 20 percent of its employees, has seen a 55 percent drop in organic search traffic over the last three years. CEO of Business Insider, Barbara Peng, explained the layoffs as precisely a massive decline in web traffic outside the company's control.

The New York Times is also experiencing sharp drops in traffic via organic search.

When the big guys run the show

But perhaps this has been a question of when, not if. It's not exactly the first time publishers have been caught off guard by new technology; the same thing happened when the internet became everyone's property and turned readers toward online newspapers instead of printed editions.

What's different this time is that a certain search giant has skin in the game. Bloomberg discovered, in a recently published internal document from Google following their anti-trust trial, that the company had considered giving news media more control over how their website data should be used in AI-driven search functions.

Then Google thought about themselves; of course they didn't want to let publishers decide for themselves whether their content should stand outside search results generated by AI. It was, after all, a field that was "developing to become an area for monetization."

Different thinking required

An increasing amount of publishers are entering collaborations with AI companies. This way, AI companies can train their models on news articles that are behind paywalls and under copyright laws. At the same time, the publishers earn money.

The New York Times has a licensing agreement in place with Amazon. At the same time, the newspaper is suing OpenAI for copyright infringement. The Atlantic went the other way and struck a deal with OpenAI. These types of agreements will only become more common, and it's only a matter of time before Norwegian newspapers like VG, Dagens Næringsliv, and Aftenposten do the same.

When newspapers experience a decline in traffic from Google, loyal readers become all the more important. The Atlantic's CEO Nicolas Thompson has doubled down on events; a place where readers of The Atlantic can meet, and the news outlet can strengthen the relationship with subscribers. Not least, the events will become an important source of income for the publisher.

We see the same tendencies here in Norway, with Dagens Næringsliv constantly inviting Norwegian business leaders to networking sessions and chatting. At a price, of course. The new gold lies in building meeting places. At least if we're to listen to Thompson and The Atlantic.

Death to the article

Generative artificial intelligence became available less than three years ago. It didn't take long before it changed the way we acquire information. Now it's up to the media houses to hold onto our attention, through brand building.

Fortunately, Norwegian editors seem to be equipped for change. During the Nordic AI in Media Summit (NAMS) in April, VG boss Gard Steiro felt that small, incremental adjustments to article format and news distribution are too passive. He believes the article's time is overdue:

The article is going to die, it should die, but storytelling is going to survive - Gard Steiro at NAMS.

And if anyone that should never be written off, it's the storytellers.

Protests and disinformation

We're staying in the media bubble. With the protests in Los Angeles comes the trouble with getting your information from AI. California Governor Gavin Newsom shared a picture of President Donald Trump's deployed National Guard sleeping on the ground. The picture was quickly sent to both ChatGPT and Elon Musk's Grok for verification. But both language models claimed the picture was from Afghanistan in 2021. This is just another example of how AI bots are not to be trusted when it comes to verifying images. You can - and should - read the whole thing over at WIRED.

San Francisco Chronicle, which first published the image that Newsom shared, has confirmed that the pictures are real and authentic.

Wikipedia did "a Google"

In a piece by 404Media we can read about Wikipedia having flirted with AI-generated summaries on their pages. This has not been met with enthusiasm by the dedicated volunteers that edit and update the Wikipedia pages. Thus Wikimedia, which operates Wikipedia, had to withdraw the service. No AI for you. At least not when visiting Wikipedia.