I changed the spelling for predictathon because I like it this way more. No further questions!

No asking “What will AI do to the web?” is too open a question for our predictathon — we really need yes or no questions, so let’s rephrase it.

Will AI destroy the web as we know it?

Short Answer: Yes

Why: Most LLMs are, at their hearts, plagiarism machines. In order to be created they required basically everything on the internet (regardless of copyright status or paywall), along with almost every book ever written (again, regardless of copyright status) — and they weren’t shy about it. Meta famously just pirated books by the boatload, and basically every AI uses “The Common Crawl” as a source — a download of most of the internet, without the author’s permission.

What that means is that, when you ask an AI “How do I change a sparkplug?” and it confidently replies “Unscrew the old one and screw in the new one,” well, that knowledge didn’t just come from the aether. Enough people wrote about it online that it created a mathematical pattern in the AI’s vector database that it could use to generate the response.

In the old internet, the PRE-AI internet, asking that question would take you somewhere that answered it, and that somewhere had an opportunity to sell you stuff — to have a sponsored video, to show ads, or even to sell spark plugs. That’s what made the internet economy work — they sold you answers and, in return, you saw ads in one form or another.

But AI is a middleman who cuts out the, uh … end man? I don’t know what you call the person on the other side of the middleman, but they don’t get traffic anymore. The traffic stays at the AI — and they can show you ads, or offer to sell you stuff, or do whatever they want. But what they don’t do is send you off to someone else to answer your question.

The sad part is, the answer still relies on other people’s efforts, they just aren’t paid anymore. Let’s use an example: AI powered News Aggregators (the only Kagi feature that I really, really don’t like).

An AI powered news aggregator scrapes news from other sources — typically real news rooms (or other aggregators). These real news rooms tend to pay people to go report on the news. They might pay for photographers to go to a location and take pictures, or purchase pictures from someone there. They pay people to do research and generate copy. They pay editors to improve it. They pay fact checkers to make sure it’s accurate. All of that is overhead required to produce news.

An aggregator takes all that work, and then just rewords the text and posts it on their page. Sometimes they link to their sources — often they don’t. Then they can run ads, or charge a subscription, or make money in some other way. They make money, the organization doing the work no longer does.

You can see how this story ends. Starved of revenue, news sites will go under — the more they are spending to generate good news (hiring reporters, fact checkers, editors, etc.) the FASTER they will go under because their overhead is so much higher. That means lower-quality newsrooms will last longer than high quality ones. The quality of the news available to everyone will slowly degrade over time — eventually we will be left only with news sources that are backed by wealthy groups with incentives to subsidize newsrooms — for example, governments that want to tell a certain story (or not tell it).

It’s bad! It’s real bad news, folks!

And that is what will happen to every kind of media that relies on ads.

NOW! Some sites may survive — those that can prove their worth sufficiently to generate a dedicated, paying audience. It’s not impossible (Defector.com is one good example), but it’s very, very hard. But a lot of sites, especially enthusiast sites that have niche markets, will wither and die.

We can extrapolate out how this will be perceived by most people. As these sources of information die, LLMs will essentially become frozen in time — with nothing new to train on, they will continue to return results based on the state of the a niche when the reporting or community around that niche gave up the ghost.

When that happens either people will notice, and seek out new sources of information, and new opportunities will arise, OR they will just deal with the slightly worse, slightly stale search results. Which do you think it will be?

Either way, the web as we know it — as it’s functioned for the past 25 years or so — will be dead.

Will it be replaced by something better? I don’t know. It will be replaced by something more centralized, something that is more of a black-box, and something that is guided by a few extremely powerful AI companies. I try not to be an AI doomer, I try to just take the middle road and admit that there are both good and bad things about AI, but this is something that legitimately bothers me. Probably why my first commandment of AI usage is what it is.

Because it could be scary. Or we could take action and fix it — by paying for news that matters, or just by not using AI for search. It’s actually not that terribly complicated.