Build Log For game devs

Crit Pick, month two: one page per Pal, and the web that ties them together

A historical Crit Pick build log about page structure and internal links, with the planned mechanisms kept separate from measured results.

Month two, and Crit Pick still gets almost no traffic. The honest number is still tiny, and I'll put the real figures to it in the next checkpoint.

This post is about the thing underneath the numbers: the shape of the site. Over the last few weeks I've been splitting Crit Pick into a lot of small pages and wiring them together, and the whole bet is that the structure earns rankings before a brand-new site has any authority to spend. Here's how the pages are split, why one Pal gets one page, how they link, and the bets I'm making that the next post gets to grade.

Crit Pick is an SEO case study I run in the open for SeeIndie, and also a games project I'd wanted to build for years. If the method works, you'll see it here in real numbers. If it doesn't, you'll read that too.

Month two checkpoint · 18 August 2026

The short version

The unit is the page, and the page follows the search. One Pal gets one page, because people look up one creature at a time. Volume per page is small on purpose.
A page only exists when demand splits to it. Otherwise it's a section inside a bigger page. Thin pages that fight each other are the failure mode.
The links are the actual product, not the pages. A pillar routes down to a hub, the hub to the individual pages, and those link back up and sideways to each other. A page with no links in and out ranks for nothing.
A link a crawler can't follow isn't a link. My navigation was built in JavaScript, which the AI crawlers never run, so in the raw HTML the whole site was unreachable. That got fixed.
The bet is structure first, authority later. It's slow, the numbers are small, and the next checkpoint is where you get to grade whether the shape actually paid off.

The unit is the page, and the page follows the search

The first decision on any topic isn't what to write, it's how finely to cut it. Get that wrong and you either bury ten searches inside one page nobody can find for any of them, or you spread one search across ten thin pages that compete with each other.

My design choice was to give individual creatures a page when that page could answer a distinct need. Other readers want a complete list or a comparison. Those are different tasks; neither format is automatically the right SEO choice.

The opposite case is RimWorld mods. There are dozens of Vanilla Expanded modules, and most of them don't have enough search demand to earn a page of their own. Those become sections inside the parent page. A module only graduates to its own deep dive when the data shows people searching for it by name. A "best of" ranking is a different animal again: a tier list targets people comparing options, so it's its own page, and it links back down to the deep dives it ranks.

None of that is clever. It's just refusing to let the site's shape drift away from the shape of the search.

Palworld, in three layers

Here's the whole idea in one worked example, because Palworld is where I've pushed it furthest.

The pillar

The Palworld hub. It introduces the topic and routes you deeper. One page, broad.

The hub

The Paldex: a grid where every Pal is a tile, and every tile links to that Pal's page. 34 of a planned 287, and it says so plainly.

The spokes

The Pal pages themselves, one per creature. Each links back up to the hub and sideways to its neighbours.

The top layer is the pillar, the Palworld hub. It exists to introduce the topic and route you deeper. Under it sits the Paldex, a grid page where every Pal is a tile, and every tile links to that Pal's page. Right now the grid holds 34 of a planned 287, and it says so plainly, because a page that claims to be complete when it isn't is the kind of thing this whole project exists to not do. The bottom layer is the Pal pages themselves, one per creature.

Then the links start doing the work. A guide I'm about to publish, the best early-game Pals, ranks nineteen of them from strong to situational, and every one of those nineteen names links straight to its Pal page. That's a guide pouring visitors and link value down into the individual pages. The pages link back up to the Paldex and the pillar, and they link sideways to each other wherever it's genuinely useful, a breeding chain that needs two parents, a prev-and-next walk through the dex. No page is an island, and none of the links are decoration. Each one is a path a reader, or a crawler, can actually take.

The links are the product, not the pages

The structure helps readers and crawlers move between related pages. Whether it improves rankings is a separate question; this build log does not isolate that effect.

The animal guide is an observation worth tracking, not proof of how link value works. Its early figures cannot separate the effect of structure from topic, competition, time or other changes.

That's why I keep saying the links are the product. A pile of good pages with nothing connecting them is just a pile. The cluster is what turns thirty small pages into something Google reads as an authority on a topic.

A link a crawler can't see isn't a link

One hard lesson belongs here, because it nearly sank the whole plan quietly. My site navigation, the menu that reaches every hub, was built in JavaScript. It looked perfect in a browser. But the AI crawlers, the ones ChatGPT and Perplexity and Claude send out, don't run JavaScript at all. So when I checked the raw HTML, the version those crawlers actually read, the site had 19 internal links and reached zero of its nine hubs. The entire structure I'd just described was invisible to exactly the crawlers I write about.

19 → 32

internal links a crawler can follow in the raw HTML, on every page

Crit Pick build log

0 → 9

of nine hubs reachable without JavaScript

Crit Pick build log

The fix moved the hub links into the part of the page that ships as plain HTML. Raw-HTML internal links went from 19 to 32, and reachable hubs from none to all nine, on every page. The lesson is blunt and it's the same one SeeIndie keeps making: a link that only exists after JavaScript runs is a link half the web can't follow. If it's meant to carry ranking, it has to be in the delivered HTML.

The honest part: this might be thin content

Here's the risk I'm carrying, said out loud. Thirty-odd near-identical Pal pages, on a brand-new domain with no authority, in a space that established sites have owned for years. There's a real reading of that where Google sees a stack of thin, samey pages and treats the lot as low value. I don't get to know yet which way it goes. The pages are new, they have no authority to lean on, and most of them are sitting on page two or three doing nothing.

Worked example, not a promise. I made peace with the trade anyway, for two reasons. One, if per-creature search is real, then per-creature pages are the honest way to cover it, and the alternative, one giant list, serves the reader worse. Two, and I'll be straight about this, some of these pages exist because I wanted them to. Crit Pick is a games project I actually care about, not just a test rig. A page built out of real interest is never fully optimised, and I think that's a feature. A site that only ever ships the keyword-perfect page reads like one. Whether the bet was smart is a question the data answers, not me, and that's the next post.

The bets, so you can grade me

That's the point of doing this in the open, so here are the open questions I'm handing to the next checkpoint:

Does one page per Pal earn its rankings, or does the cluster read as thin content? Does the link value actually flow up to the hubs, or do the small pages rank while the hubs stall on page two? Does the evergreen tier list keep compounding? And does the one high-volume utility page I bet on, the event timer, ever climb once the cluster around it builds enough authority to carry it?

I don't know the answers. That's the honest state of it at month two. The structure is built and the bets are placed. Next time I'll put real Search Console numbers to every one of them, what climbed, what stalled, and what I got wrong.

Common questions

Isn't one page per Pal just thin content?

It might be, and I've said so above. The defence is that the search genuinely splits per creature, so per-creature pages match real intent rather than padding. The risk is that a new domain with no authority gets read as low-value regardless. The data decides, and I'll report it either way.

Why not one big Paldex page instead of hundreds of small ones?

Because almost nobody searches for the whole dex. They search for one Pal. A single giant page can't rank well for two hundred different creature names at once, and it makes a worse reading experience for the person who wanted just one. The grid page exists to tie them together, not to replace them.

Does internal linking really do anything this early?

It's the clearest positive signal I have so far. The one page that grew without me touching it grew because the cluster around it filled in and the internal links accumulated. On a site with no backlinks, internal structure is most of what you've got.

Can I check any of this?

That's the whole idea. Crit Pick is live at critpick.me, every page here is public, and the numbers in the next post come straight from Google Search Console. If a claim here doesn't match what you can see, I want to know.

Sources

Google Search Console (clicks, impressions, position) — via the Crit Pick case study, data through 5 July 2026
Crit Pick content map and build log — the site architecture and the 19→32 internal-link / 0→9 hub counts in the raw HTML
Own keyword research — the event timer at roughly 12,100 searches a month