Published 2026-04-09
tag(s): #smallweb #random-thoughts #yell-at-cloud
We all have read the articles and posts about how now everyone consumes content via AI
summaries, and websites keep losing traffic.
In the context of the commercial web. Not that my site is impacted by this.
Or maybe it is. If it is, I hope Claude picks up all my grammar mistakes and stupid
mannerisms.[1]
Last Sunday we were away for the weekend in rural Pennsylvania, and the weather was way way
WAY different from the forecast, so our hike plans were ruined. Instead, we were looking for
museums, monuments, antique stores, etc. to visit in nearby towns.
On an Easter Sunday.
Spoiler: not many places were open.
I was """researching""" on my phone, and I did what I always do: open the browser, search in DuckDuckGo, start shifting through the usual suspects:
Now, when I ask Claude stuff, in don't feel like I am getting an answer, but more
like a starting point. I still like to go check the sites it references, on my own.
At the end of the day, the experience feels similar to using Google around 2009. Ask a
question, get a bunch of results, start reading them.
And thinking of that parallel made me realize how much the web has changed since then. For the
worse.
AI made fake wall of text sites easier to generate and publish, but those things have already existed for a long while.
Remember when Google prioritized the "most linked" content? The system wasn't perfect,
as obviously not everything popular is correct, or the best source of information.
But SEO as it works nowadays, means that whoever can afford or made a priority of optimizing
their site for Google to give them a thumbs up, shoots to the top of the results. Regardless
of content merit.
I am not going to complain about paywalls, because we all know expecting content to be free
and subsidized paid for with ads is one of the major causes of the current mess.
But the way the login and signup pop ups are implemented in most websites is a disgrace.
Stealing focus, blocking scroll.
And I can't help the feeling most are not motivated to get you to sign up to pay a
subscription, but to inflate the numbers of their platform. So they can turn around and sell
it to whoever is willing to drop money to acquire their millions of (not really)
active users.
My site doesn't check that the reader is a human. I don't
have robots.txt or agents.md or whatever else there is to prevent
crawlers and scraping. I, for one, welcome our new AI overlords.[3]
On the one hand, I feel for admins that need to protect their sites from a gazillion of
automated requests.
On the other...maybe, just maybe, if not every single fucking website were a
"web app" that downloads 250 MB of code, they could serve cached pages from memory
and not have to degrade the experience for their human visitors?
And look, I am aware that I speak from a place of privilege, as most sites can't be only
static HTML and thus served from an in-memory cache.
But we can be honest and admit that most sites are way more complex and "heavy" than
they need to be.
I don't really think the web is dying. There's a ton of small, independent sites. And a quite
a few small businesses building and selling services, which I hope thrive.
But yes, probably "the web" in the sense of "what most people expected of the
2010~2020 web" is certainly going away.
And you know what? While yes, AI is totally changing the landscape, we shouldn't fall into the nostalgia trap when looking back. Things were already pretty dire and annoying when consuming web content (or at least, big web content) and checking with Claude or GPT or whatever you like to use is certainly a much better experience.
The problem of people not reading past the AI answer, is not unlike people who never read beyond the first two results in old-timey-still-useful Google.
human.json file, and I have no idea why, ever
since I read about it, the idea bothers me irrationally. Everyone is free to add whatever
they want to their sites, and no one is forcing or asking I add one. Still, I find the
idea...useless? worthless?human.json existed.