Traffic, referrals and demand for front-end technical writing have been falling for a while — CSS-Tricks' own numbers, and Stack Overflow's, tell the same story, and other content creators and publications report it too. Holding the author personally responsible is tempting, but the root cause sits mostly beyond individual control: AI is the biggest headwind, even if it isn't the only one. The subject worth examining is technical writing in the age of AI.
Who the writing is still for
AI does not learn on its own; it only goes as far as its prompt, and human curiosity and effort are what push things forward. AI is not the new primary audience for technical writing: the motivated people who want to learn keep the practice alive, even when that learning later becomes the prompt from which AI generates code.
Where technical content still fits
Asking who — or what — a piece is written for may reveal that your place in technical content needs to shift. Technical documentation is probably worth less than it once was: the specs and MDN already hold a wealth of it, and developer questions of that kind now get answered in a chat embedded directly in the IDE.
CSS-Tricks, despite being full of that material, has developed a clear role of its own: making technical documentation more accessible to humans. The Almanac has been grown and maintained since around 2009 and was expanded a couple of years ago. The specs are dense and deliberately precise so that browsers can implement CSS features; the site aims to offer human explanations for technical topics, closer to talking with another developer over coffee, and to lower the barriers to access.
It also works as a platform for ideas. The site began in 2007 as Chris Coyier's personal learning space and changed once guest authors were welcomed; 748 authors have contributed since, the author of the piece included. Plenty of those tips and tricks might never have appeared without a place to publish them.
Deciding whether to keep going
Whether writing is still worth it has no universal answer. Money, popularity, craft and personal learning — or some mix of them — are all valid reasons to continue or stop. It may be financially unsustainable to run a blog or newsletter at scale, other demands may compete for your time, or a good new job may be more appealing. Or you may simply be burned out, given the time, effort and expectations involved and a payoff that may not measure up; the result then is apathy, and some days do feel worse than others.
Still, there is a place to fill, and real fulfillment in a comment, email or mention confirming the work helps someone. Telling your favourite technical writers that you rely on their work costs nothing.
A pendulum, not an ending
This is not a prediction that AI collapses and everyone returns to writing code entirely by hand. Technical writing, and creative web design as a whole, has been in this position before. Jay Hoffman's history of web design records a time when sharing what we learn ran against the grain: articles on Word were one-of-a-kind, their images, backgrounds and colours carrying the tone of a piece, art-directed in a style blending 8-bit graphics and off-kilter layouts with the chaotic bricolage of punk rock zines, so that pages came alive and expressed the personality of their author.
Articles that appeared on Word were one-of a kind, where the images, backgrounds and colors chosen helped facilitate the tone of a piece. These art-directed posts pulled from Levy’s signature style, a blend of 8-bit graphics and off-kilter layouts, with the chaotic bricolage of punk rock zines. Pages came alive, representing through design the personality of the post’s author.
Web design was punk rock in the 1990s and early 2000s, then became a discipline, and is now becoming automated — which makes human web design punk rock once again, or at least gives it the chance to be.
My apathy levels for the industry are absolutely sky high and I don’t even have a dickhead boss making me hit token targets
— Andy Bell (@bell.bz) 2026-05-19T18:23:42.034Z
More and more people I deeply respect and have learned a lot from over the years feel like they are speaking into a void, to the point where they are losing motivation to continue making content, which got me to write my first post on my site in over a year…www.kevinpowell.co/article/tell…
— Kevin Powell (@kevinpowell.co) 2026-05-21T13:22:13.240Z
It's not fun to make tech content as much anymore, too. I'm doing some because I actually like it, but it feels like I'm shouting into a slop-driven void nearly everywhere. It's really tiring.I get so much joy reading individual blogs and newsletters still (like yours!), at least.
— Cassidy (@cassidoo.co) 2026-05-19T22:32:55.483Z
I’m just 7 videos away from finishing a 45-video course about building for the web, and the fact that people are rarely choosing to learn from human-crafted content anymore therefore rendering hundreds of hours of my human effort completely pointless is burning me out catastrophically. I am a husk.
— ⭑ salma (@whitep4nth3r.com) 2026-05-20T07:05:14.119Z
Where AI fits — and where it doesn't
The 2019 advice holds up; nothing in it needs subtracting. But the landscape has shifted since, and a few additions are warranted. The first is narrower than it sounds: be cautious about using AI to write.
Two problems stand out. Language models are not always accurate, and their output dilutes your personal voice — both bad for the discipline. And there's little point handing readers an AI explanation of something they can pull from their IDE in seconds. That is exactly how AI slop happens.
The caveat: this applies to the writing itself. Grammarly was useful long before it was marketed as a full AI writing suite, and the beefier features that practically do the writing for you aren't required to get value from it. AI is a reasonable reach for low-lift work — quick spell checks, converting Markdown to HTML, keeping the publishing calendar on schedule. Much of what gets called "AI" today is what we used to call "automation"; the interesting tools enhance the craft rather than replace it.
Write what a model can't supply
Basic definitions and quick code examples for a CSS property are well-trodden ground — a model produces them easily, often pulled straight from existing documentation. Unless you're repackaging that material differently, aim for what AI can't solve on its own or can't solve easily.
The path from novice to understanding is what makes the difference for a technical writer. Struggle is where learning happens: a client asks for something you've never attempted, you make mistakes, you work through them. That journey gives readers a mental model they can apply to their own work — even when the mental model ends up going straight into an AI prompt. It's the same instinct as the View Source generation, or musicians slowing down a record player to learn chords, notes, and beats.
The authority question
There's an expectation that everything published on a site like CSS-Tricks is the "right" way and that snippets can be copy-pasted as-is. That promise was never made. CSS is poetic in the sense that more than one approach works, and the best one fits your mental model most snugly. Trying something purely for the sake of trying it has real value.
Exploring, experimenting, and sharing along the way sometimes means sharing just the germ of an idea — no dogma required. The goal is to be experienced, not cynical. You don't need to be the be-all-end-all authority on a subject to earn permission to share; experience and expertise aren't always the same thing.
Learning in public, and citing the people who taught you
What you learn will mean more to a real person than some generated replacement for a Stack Overflow answer. All of it is worth writing up:
- A thing you tried that didn't work out.
- A thing you tried that did work out.
- The mistakes made on the way to getting it right.
- Open questions you still have about what you learned.
That's the source of real-life experience and of learning paths you simply don't get from generated responses claiming to have the best answer. Solve real problems, get into the edge cases, leverage community knowledge.
None of us wakes up with all the answers; we learn from what others learn and share. Point readers in their direction. Wanting to appear as the original source of everything you publish is tempting, but it's essentially never true. We build off each other — that's blogging at its core — so build each other up the old-fashioned way, with hyperlinks.
Traffic, SEO, and AIO
CSS-Tricks toe-dipped into SEO years ago, but it was never really the focus. Riding the organic search train without worrying about keyword stuffing, click-baity headlines, or anything else optimization-shaped has been liberating; the effort goes into writing for humans, good structure, and a consistent voice and tone — the things Google used to care about, and the core value of which holds less value than ever even if Google still claims otherwise. Search traffic is being gutted in favor of generated answers tied to the services people use, which is a clear hit to traffic and to the revenue models that traffic analytics measure.
"Artificial Intelligence Optimization" (AIO) is best forgotten for now. The "best" practices are fluid and inconsistent — some agents favor proposals for text-based files to help LLMs use a site, while others, Google among them, aren't interested. It may mature into a discipline with clear strategies and guidelines; we don't know yet. Feigning ignorance until the smoke clears is a defensible position, and whether it even matters for CSS-Tricks to surface as a reference in a generated response is an open question.



