I published 263 articles in ten days. Here's what Search Console showed afterwards.

Between 30 July and 8 August I put 263 pages live on Crit Pick, my games site. A hundred and four of them went up on a single day. Five weeks later the site's average position had fallen from 18.6 to 32.3 and its clicks had gone from sixteen a month to four. This site, SeeIndie, published 34 articles in the same period and got 29 clicks in the same 28 days. Per article that's a factor of 84 in favour of the smaller site.

That's the headline, and the honest thing to say next is that I can't fully explain it. What follows is the whole file: what I published, what the numbers did, the two times I read my own data wrong before I caught it, the confounder I can't subtract, the one thing the data does support — and a cost that none of the metrics show at all.

Data through 6 September 2026 · Google Search Console, both properties · Crit Pick is an SEO case study I run in the open for SeeIndie. Every page named here is live and every number is one you can check.

What I actually published

The sprint ran 30 July to 8 August 2026. Ten days, 263 pages. I've said "two weeks" about this before and that was loose: across a full fourteen-day window it's 283, and the tight ten-day count is 263. Both numbers come out of Ghost's admin API and out of my own publishing log, and the two agree.

The shape matters more than the total. The biggest day was 8 August with 104 pages. The second was 4 August with 54. And 256 of the 263 carried one tag: Palworld. 254 of those were Paldex entries, one page per creature, generated from a data set and published in batches.

So this was never "I wrote a lot". It was a database that happened to be shaped like a blog. My rule was, and still is, that every creature gets its own page, because that's how people search: one creature at a time. What I was testing without quite admitting it was whether a two-month-old domain could publish a complete reference for a game and have Google treat it as one.

Before the sprint the site had 90 published pages. By 6 September it had 393.

What Search Console showed

Both of my sites over the same 28 days, 10 August to 6 September.

Same person, same platform, both started in June 2026. The site with a twelfth of the stock takes seven times the clicks.

Crit Pick didn't only underperform the other site, it underperformed itself. Across three months, 7 June to 6 September, it averaged 16 clicks and 2,553 impressions a month at position 18.6. In the last 28 days of that same stretch: 4 clicks, 2,020 impressions, position 32.3. The stock more than quadrupled and all four numbers fell.

That's where I stopped the first time, and that's where I made my first mistake.

Where I nearly published something false

The comparison above sets a three-month average against a 28-day window. On the strength of it I wrote "the site is in decline". Then I pulled the daily series, blocked it into fortnights, and the shape changed.

The collapse is real. The continuing decline was not. My 28-day window happened to end at the low point and contained none of the recovery. Over the last eight days the site is seven places better and the clicks are off zero, on essentially unchanged impressions per day: 72.6 in the low window, 69.8 in the recovery one. The movement is coming from positions, not from how often the pages get shown.

Eight days and three clicks is a signal, not a result. I'm re-measuring on 23 September, 7 October and 4 November, and publishing whatever it says.

The second thing I got wrong, which I caught before publishing this

Having found the collapse, I wanted to date it. Google confirmed an August 2026 spam update starting 18 August, running 2 days and 16 hours. If the fall demonstrably started before the 18th, I could rule the update out.

So I wrote: the first drop lands in the block from 2 to 15 August, which is after my sprint and before the update. Clean.

It isn't clean, and the reason is instructive. That block still contains 217 of the sprint's 263 pages. The 2nd to the 8th of August is the sprint. A fortnightly average that spans an event can't tell you what happened after the event — and I'd just built a smaller copy of the exact window-selection error I'd caught myself in one section earlier.

Here's what survives once I cut what I can't support. The block ending 15 August ends three days before the update began, and by then the average position had already gone from 17.9 to 27.3. So whatever cost the first ten places, it wasn't the August spam update. The second stretch, 27.3 down to 36.5, overlaps the update and I can't assign it. Nor am I claiming causation from two properties, one time period and single-digit clicks: a Google update, seasonality in gaming, or something technical at my end are all still live options.

What the data does support: the field, not the volume

Strip out everything I can't defend and one finding is left, and it's the useful one.

I published into three clusters with the same method in the same window. Where they sit now, from Ubersuggest on 9 September:

Not one Palworld page ranks above 55th, and each ranks for exactly one term. The RimWorld guide ranks for eight and reaches 16th.

The Guild Wars 2 page makes the same point without changing the page at all. On "which race should you play", the long-tail terms — 30 searches a month, difficulty in the low twenties — sit at 27th. The head terms on that same page, "guild wars 2 races" at 1,900 a month, sit at 59th and 60th. One page, one crawl, two very different fields.

There's a second piece of evidence for this that has nothing to do with Search Console, and I like it better because it comes from a different tool on a different day. When I run a content tool over one of these database terms, the competing results it finds are stores, databases and news items — no prose at all. That's the field. Paldex terms belong to established game wikis with years of links behind them. A three-month-old domain with one referring domain doesn't take that ground by turning up 254 times.

Two limits on it, honestly. Four Palworld pages out of 254 is a sample, and not a random one: those four are the only Palworld pages that appear in Ubersuggest's top-pages list for the domain at all, so I checked the most visible ones and they're still on page six. And the tool's coverage is thin at this size — it finds 74 ranking keywords for the whole 393-page site. Strong hint, not proof.

What Google actually says about "Discovered – currently not indexed"

If you read one section, this one.

Crit Pick's index report shows 51 pages under "Discovered – currently not indexed", plus five under "Crawled – currently not indexed" where validation has failed. Every SEO write-up I've seen reads a number like that the same way, and so did my own first draft of this post: Google knows about these URLs, Google has decided they aren't worth having, this is what a site gets for publishing at scale without the authority to back it.

I went to source that claim in Google's documentation, and it isn't there.

Google's definition of the status is: "The page was found by Google, but not crawled yet. Typically, Google wanted to crawl the URL but this was expected to overload the site; therefore Google rescheduled the crawl." That's about serving capacity. It says nothing about quality and nothing about authority. The wording is in the Search Console page indexing report documentation.

The crawl budget documentation doesn't rescue the folklore either. Google defines crawl budget as "the set of URLs that Google can and wants to crawl", and says the allocation factors in "things like popularity, overall user value, content uniqueness, and serving capacity". Popularity and uniqueness are in there. The words "quality" and "authority" are not.

I'm not telling you the folklore is wrong. I'm telling you it isn't documented, that I repeated it before I checked, and that those are different things. I can report the state of my index report. I can't tell you what Google meant by it.

Two more things belong beside that number before anyone quotes it, me included. Of the 124 pages Google isn't indexing, 18 are plain 404s and five are blocked by my own robots.txt — my mess, unrelated to publishing speed, still not cleaned up. And the report covers 492 URLs against 393 published articles, so roughly 99 of them are presumably tag, author and pagination URLs that Ghost generates on its own. Until I've separated those out, "51 pages" isn't purely a content number.

While I was in the documentation I read the spam policy on scaled content abuse too, since that's the other thing people reach for here. Google's definition: "many pages are generated for the primary purpose of manipulating search rankings and not helping users ... large amounts of unoriginal content that provides little to no value to users, no matter how it's created." The hinge is purpose, not quantity. I have no manual action and nothing pointing at that policy, so I'm naming it as a hypothesis I checked and set aside, not as a finding.

The bill that no metric showed me

Here's the part I didn't expect to be writing, and it's the one I'd act on first.

Palworld's 1.0 release renumbered its entire creature index. That change appeared in no line of the official patch notes — it took five independent sources before the finding was solid enough to trust.

My database pages carry those numbers in their titles. "Astralym — Palworld Paldeck No. 204". I counted this week: 288 live pages on the site have a Paldeck number in the title.

That's the real cost of publishing a database as a blog, and it's invisible in every number further up this page. Search Console will never tell you that your titles have gone stale. It has no opinion about it at all.

The question I haven't answered

There's a hole in everything above and I'd rather point at it than have you find it.

Every number I've used for the decline is a property average. A property average can fall for two entirely different reasons: because the pages that were ranking dropped, or because 256 new pages entered the index below position 55 and pulled the mean down while nothing else moved. The second is arithmetic, not a judgement from Google, and I've been treating them as one thing.

The clue I have points the wrong way for me, and I can't use it cleanly. RimWorld animals, the best page on the site, had 992 impressions and 16 clicks in the eight-week cumulative view through 29 July. In the 28 days to 6 September: 203 impressions, zero clicks. Dilution doesn't take clicks off a page that already had them. But one of those is a cumulative view and the other a 28-day window, which is precisely the comparison I spent two sections warning you about, so it's filed as suggestive and nothing more.

Settling it needs the daily series for individual pages, not the property total. That's the pull I'm doing on 23 September, and it's the first thing in the next post.

What I'm doing instead

Three things, and none of them is "write more articles".

Cash in the visibility I already have. The site collects 2,020 impressions a month and converts four. Four pages hold 573 impressions between them and zero clicks — already being served, just too deep. Going from position 32 to position 5 is worth roughly 25 times the current traffic without a single new page. For the version of this problem with no sprint attached, I wrote it up in "Why your website isn't showing up on Google".

Fix the ratio that caps everything else. One referring domain against 393 articles. SeeIndie has eleven against 34, and ranks eleven places better. I've spent three months making pages and no time making the case for anyone to link to one, which is an odd allocation for someone who writes about SEO.

Split the database off the blog, and measure it. On 16 September the generated database moves to its own subdomain, starting with a new game's creature index. The editorial blog stays where it is.

The honest verdict

Ten days, 263 pages, and the sharpest thing I can hand you is a caveat: the site got worse on every metric, the trend I first reported was an artefact of how I sliced the data, and half the decline sits on top of a Google update I can't subtract.

What's left is still worth the ten days. Volume aimed at a field owned by established sites produced 254 pages ranking for one term each, none above position 55. The same method aimed at a field nobody had fully answered produced a guide sitting 16th. The variable was never how much I published. It was where — and, as it turns out, what I'd be on the hook to maintain afterwards.

Four clicks a month isn't a disaster. It's a data point about a three-month-old domain, which is exactly the age at which everyone else quietly stops publishing their numbers.

What's next

Re-measurements on 23 September, 7 October and 4 November, published here either way. Next time: the page-level daily series that says whether the old pages fell or the new pages just diluted the average, whether the recovery held, and an answer on whether those 288 titles carry the current numbering. In December, the subdomain result.

Common questions

Sources: Google Search Console (clicks, impressions, CTR, average position) for critpick.me and seeindie.com, 28-day view to 6 September 2026 plus the 90-day daily series; Ghost admin API for the publication counts, read 9 September 2026; Ubersuggest for the cluster positions, 9 September 2026; Google Search Console Help (page indexing report), Google Search Central (crawl budget, spam policies) and the Google Search Status Dashboard for the August 2026 spam update dates.