Ben Daubney

The web used to be a reliable library. AI has ruined it.

This post is part of a series for #WeblogPoMo2025. Read the introduction here.

Have you tried searching for help on the web recently and seen how depressingly useless it has become?

Here's a real example.

I moved into a new home recently and got given a huge folder of instruction manuals and warranties and paperwork. Instead of scouting through them all, I figured that I could quickly look online to figure out how to adjust the temperature in the refrigerator.

I went over to DuckDuckGo and entered my search term: "how to adjust temperature on my fridge".

The first page of results gave, in order:

I gave up, dug through my box of warranty cards and wiring diagrams, and found the solution offline.

I vividly remember first using the web. It was November 1995, our school was just starting to get their hodgepodge of donated PCs networked, and if you asked nicely you could get online sometimes. I drank in everything I could, jumping from incredible early screenshots of Mario 64 to descriptions of Father Ted episodes I hadn't seen.

Everything was there! It was incredible! There was so much real information, all freely given and freely available!

The web-as-reference-guide has been set in my mind ever since. Part Hitchhiker's Guide, part directory of all human interest, it's the thing about computing I've come to love the most. I've never been a big social networking guy and email is great and all, but the best of the internet for me has always been the comprehensive and esoteric and authoritative World Wide Web.

In 2025 the web isn't the main use of the Internet any more. Apps are a far bigger deal, walled gardens of social networks even bigger. Knowledge isn't as valued a commodity as updates. Get people in, get them hooked, get them coming back for more.

No matter how much the site begs, I'm never going to sign up for updates from TheKitchenApplianceDad.com.

Good websites take time. Inspiration on what to include, time to draft the content, more time to refine that content. Blood, sweat, tears. Sometimes joy. Always work.

How often in 2025 does someone think:

"Gee, I know a lot about the controls for refrigeration so I'll sit down and write a cogent, simple, universally applicable article that's easily found online"

?

It takes a particular individual to consider doing that work for little if any reward.

Maybe you're the sort of person who might write an article if there's some money in it. You could design something with Amazon affiliate links, maybe some Google AdWords or those Google AdSense placed adverts. Your page on fridge settings isn't going to set the world on fire so most likely you'll get a few pennies here and there.

From that point of view the time and effort isn't worth the monetary reward. You either write for the sheer hell of it or you walk away and do something better with your time.

Generative AI completely upends this line of thinking.

Literally anyone can very quickly:

In an hour it's very easy to have a brand new site up and running with twenty articles all generating a little bit of income for next to no monetary or intellectual outlay.

If you're technically minded it's possible to automate a lot of these steps, to ask the AI tool to generate a list of potential posts and then to automatically take each item on that list and write a post.

Depressingly easy.

The content is depressingly empty of meaning too. All have the same structure: an introductory paragraph, some key considerations, a list of ambiguously generic steps to 'solve' your problem, some key takeaways and a conclusion.

There are no specifics, just generalised vibes along the lines of:

"You probably have a dial that you need to locate so find that and then usually the bigger number on it is the warmer setting but it might not be so maybe experiment a bit".

You're still none the wiser but you've clicked on the site so the 'author' has got a bit of Google AdSense revenue. Bully for them.

I see this same pattern for all sorts of non-specific, awkward technical problems where the web would've produced a definitive answer a few years ago.

An aunt had an ambiguous iPhone problem last year, something about her Apple account not set up properly, the sort of difficult issue that you hope someone else has had and walked through how they solved it. I tried to help my aunt but every search I made led me to AI generated nonsense.

I'm putting up shelving in my garage. What should I be considering, particularly in the UK where garages will be prone to changes in the weather? Dozens of pseudo-articles trying to get me to purchase particular drills or insulation or screws, little from people actually chatting about what they've done.

I hear a podcast referencing a fascinating anecdote about how the microchip shortage during the covid lockdown led directly to a scarcity of gummy bears. Every search takes me to conspiracy websites and meaningless empty analysis of the impact of the lockdowns.

There's no simple way to filter out these results in DuckDuckGo or Kagi or Google or whatever your search engine of choice may be.

...and maybe that's the problem:

AI-generated nonsense is appealing to search engines, which means human-generated content is more difficult to find than ever.

Good content is there, but the good content is hard to find. A new solution is required.

#AI #WeblogPoMo2025 #main #technology