Building a Guild Wars 3 page before the game exists

The Guild Wars 3 page has two jobs that pull in slightly different directions: rank for the head term “guild wars 3,” and work as a living news tracker someone bookmarks. Almost every decision came down to serving both at once. Here are the working notes, including what I still need to fix.

The job I was actually solving

Job one: rank for “guild wars 3.” That's the 5,400-volume head term, and this is the pillar built to win it. Job two: be a living news tracker that someone bookmarks and comes back to every time ArenaNet says something, which is why I eventually put “News” in the title.

A pillar wants depth and a stable, answer-first structure. A news tracker wants the freshest thing up top and a page that doesn't feel like a chore to re-scan. Those aren't opposites, but they're not the same page either, and the whole design is me trying to land both without compromising either.

Why this one is genuinely hard

I want to be honest about the constraint underneath all of this: optimizing a page for a game that was announced three days ago and won't ship until 2028 is awkward, and no amount of on-page polish fully fixes it.

The facts are thin and they move. Almost everything people will eventually search (classes, builds, the meta, “is it good,” a review) doesn't exist yet, so I can't write the deep content that normally earns a ranking. What I can write is a tracker, and a tracker is by definition a moving target: I'm building structure that has to survive the facts changing under it, and re-editing it every time ArenaNet opens its mouth.

The search intent is lopsided. Right now it's almost all “is it real / when does it come out / what is it”: announcement-wave curiosity, not the durable how-to and build intent that carries a game page long term. That demand will swing hard the moment real gameplay drops, and the page has to be ready to swing with it.

And the single strongest SEO signal is off the table. Google rewards first-hand experience more than anything else right now, and it's the one thing I cannot fake or borrow: I can't play the game, can't screenshot my own runs, can't out-experience anyone. Every site covering this is working from the same press kit. Until the beta, the page competes on structure, accuracy, and speed, not on experience, because nobody has any.

So a lot of “optimizing” this page is actually restraint. Not padding it with content I'd have to invent. Not borrowing Guild Wars 2's classes to look more complete. Not chasing keywords the game can't honestly answer yet. The bet is narrow and deliberate: be early while the keyword difficulty is low and the field is empty, be the most accurate and best-structured version of the little that's known, keep it relentlessly updated, and let first-mover authority compound, then thicken it into a real pillar as ArenaNet reveals more. The discipline is the strategy. The hard part isn't doing more; it's resisting the urge to.

What I changed

I gave the page real substance, not just a changelog. The first version was mostly a date-stamped log. I added a “What's actually confirmed” section that states the things people actually search: the setting (Orr, 1,200 years before GW1, just after a god is deposed), the Seeker mount, the momentum-based movement, the action-RPG-meets-buildcraft combat, the no-sub/no-battle-pass stance, each as a self-contained block. I also added a “what's not confirmed yet” block, because honestly answering “no classes announced, no date, no Xbox” is itself the answer to a pile of searches.

I treated the IGN interview as news, not background. Colin Johanson's IGN Live sit-down carried detail the press release didn't (the 1,200-year timeline, the deposed-god lore, the no-battle-pass line). That's a dated, sourced beat, so it got its own update entry instead of being quietly folded into the body.

I reordered the page around how people actually use it. The flow is now Hero → Trailer → Facts → Updates → What's confirmed → FAQ. The Facts sheet sits high so a first-timer gets the gist in five seconds; the Updates come next so a returning visitor hits the news fast; the long confirmed deep-dive is the tail, where length doesn't cost anything.

I made the update log page itself. It shows two entries at a time with arrow controls at both the top and the bottom, so you can advance without scrolling back up. That keeps the news section short instead of letting it bury everything below it.

I added an “On this page” TOC and an FAQ. The TOC jumps to Updates / What's confirmed / FAQ. The FAQ answers eight real queries (will there be a GW3, when's it out, what platforms, Xbox, setting, classes, subscription, is GW2 dead) as a native accordion.

I fixed wording that overstated things. “Console-first” became “launches on PC, Steam and PS5 together, simply the first Guild Wars on console,” because there's a real difference between leading on console and just also being there.

How I balance SEO, AI, and a page people actually like

The honest starting point: these three agree far more than they fight. Google's own 2026 guidance says optimizing for AI Overviews is optimizing for search. There's no separate “AI SEO.” A page that's well-structured, answer-first, and genuinely useful wins all three. So I don't build three versions of anything. I build one good page and watch the few places where the goals genuinely tension.

There are three of those tensions, and I have a rule for each.

Tension 1: answer-first (SEO/AI) vs news-first (the returning user). SEO and AI both want the core answer in the first 150 words. A news regular wants the latest drop, not a definition they've read five times. My rule: let the Facts sheet be the answer-first layer, and let navigation serve the rest. The facts table front-loads the entities (Orr, Vaelwarden, Seeker, PC/PS5, beta 2027) for the crawler and the newcomer; the TOC gets the regular to the news in one click. Nobody's underserved.

Tension 2: depth (SEO) vs a page you don't dread scrolling (UX). Comprehensive pages rank, but a wall of content is a bad experience. My rule: compress the presentation, never the content. The update log is paginated, the FAQ is an accordion, the images are sized down, but every word is still in the HTML. That's the key move: a crawler and an AI read the full DOM regardless of what's visually collapsed or paged, so I lose nothing on the machine side while the human side gets a page that breathes.

Tension 3: interactivity (UX) vs crawlability (SEO/AI). Anything that hides content behind a click or a script risks the content not being seen. My rule: progressive enhancement, always. With JavaScript off, every update shows in full and nothing breaks; the pager is an enhancement, not a gate. The FAQ uses native <details>, so there's no script dependency and no INP cost at all. The content exists first; the nice interaction is layered on top.

If I had to compress the whole philosophy into one line:

Write one honest, well-structured page; make it answer-first for the machine and the newcomer; make it news-first for the regular through navigation, not duplication; and never let a visual nicety remove a word from the HTML.

A note on AI specifically, because it's where people overthink: I don't need special schema or an “AI version.” What AI citation actually rewards is the same stuff good SEO rewards: self-contained passages it can lift, clear factual statements, named entities, question-shaped headings, and visible sourcing. The FAQ and the confirmed blocks are written exactly that way, so each paragraph makes sense pulled out of context, which is how an AI selects and quotes.

What's still wrong: my fix list

I'd rather write these down than pretend the page is finished.

Bottom line

The page does both its jobs now: it reads as a pillar to a crawler and an AI, and it reads as a fast, scannable news tracker to a person. The structure carries the SEO and AI weight; the pagination, TOC, and accordion carry the UX. What's left is mostly trust and housekeeping (self-hosted images and a byline), none of it structural, all of it on the list above.

living notes, updated as the page is.