Rejected by AdSense for a Year: I Audited My Indie Game Site and Found 12 “No Wonder” Problems

Rejected by AdSense for a Year: I Audited My Indie Game Site and Found 12 “No Wonder” Problems

13 min read
Yotsuba Games
game-dev
dev-progress
#adsense
#seo
#monetization
#browser-game
#tips

The Short Version: Still Not Approved

It has been about a year since I first submitted this site to Google AdSense for review.

The result is still rejected. The reason has been the same from day one: “Low value content.”

At the end of my earlier post about getting approved for H5 Game Ads, I wrote “The main AdSense review rejected me 😂”. A not-so-funny amount of time has passed since then, and I am still standing in the same spot.

So this is not a success story. It is a rejection story.

I do not come empty-handed, though. In November 2026 I stopped writing articles and audited the whole site from end to end. I found 12 problems that made me say “well, no wonder.” Most of them had nothing to do with the articles themselves.

If you run your own game site or portfolio site and keep getting rejected, I hope this list helps. I will keep the jargon to a minimum and explain everything as if the site were a shop.

Note: Google does not publish what the AdSense review actually looks at. This post is not “how to pass the review.” It is a record of the defects I found when I audited a site that kept failing it.

The Past Year: I Believed More Articles Would Fix It

When you are told your content is “low value,” the natural conclusion is that you need more of it. That is what I concluded, and I kept adding more.

WhenWhat I did
January 2026Rejected. At that point the site had exactly one game
January–FebruaryGrew the catalog to nine games. Added rankings, reviews, comments, and Google account linking. Rebuilt the review and ranking sections so search engines could read them
JulyPublished 313 node reference pages for PixPipeline, my pixel art tool, in both Japanese and English
AugustCleaned up duplicate entries in the sitemap

Nine times as many games and hundreds of new pages later, the approval email still had not arrived.

I also caused a couple of accidents of my own along the way.

  • The day every page broke (July): I rebuilt production without including changes that existed only on my machine and had never been committed. Code I had removed came back from the dead, and every page on the site showed an error. I rolled back in a hurry. Lesson: commit first, then build what goes to production.
  • The sitemap that pointed at your own computer (August): The redirect from the old sitemap to the new one pointed at an address that only works inside the server (localhost:8080). In effect I was telling Google, “our site map is inside your machine.” I fixed it the same day.

Far from helping the review, I suspect I was actively getting in its way.

The Turning Point: 70% of My Pages Were Not Indexed

In November I stopped adding articles and opened the “Pages” report in Google Search Console. It shows how many of your pages Google has actually put in its index.

StatusPages
Indexed (shown in search)478
Not indexed1,040

A little over 30% was indexed. The other 70% or so did not exist as far as Google Search was concerned. Of the pages that were not indexed, 874 were “Crawled – currently not indexed” or “Discovered – currently not indexed.” In plain words: “we looked, and decided not to list it” or “we know it exists, but have not looked yet.”

That was when it finally clicked. What I was missing was not more articles. It was checking whether the ones I had written were reaching anyone. A page nobody has read cannot be judged, no matter how many hundreds of them you write.

From there, with help from an AI coding assistant (Claude Code), I went through the foundations of the site one piece at a time. Here are the 12 things that turned up.

Part 1: Things That Never Reached Google (6 Problems)

1. The Sign Was Hanging on the Back Wall

Every page carries “name tags” for search engines: the page title, the description, a mark that says “this is the official address of this page” (canonical), and a mark that says “the English version is over here” (hreflang). These belong at the very top of the page, inside <head>.

This site runs on Next.js 15, which, to show pages faster, sometimes streams those name tags into the page body (<body>) afterward. And according to Google's own documentation, the canonical and hreflang marks are ignored unless they are in the head.

In shop terms: I had hung the shop sign on the back wall instead of over the entrance. Nobody walking past could tell what kind of shop it was.

The fix: One line in the config file, so that everyone always gets the name tags at the top.

// next.config.mjs
const nextConfig = {
  htmlLimitedBots: /.*/,
};

The page now waits for the name tags before it starts sending the body, so the first paint can be slightly slower. That is far better than not being indexed, so I took the trade.

How to check: Open a page in your browser, right-click, and choose “View page source.” If <title> and rel="canonical" appear before </head>, you are fine. If they are inside <body>, you have the same problem.

2. The Curtains Were Drawn

A site can post a notice called robots.txt that tells search engines “please do not look here.” Mine listed a path called /_next/.

/_next/ is where Next.js keeps the parts that make up a page's look (CSS) and behavior (JavaScript). Block it, and Google receives only the text and judges a broken page with no styling and no layout.

In shop terms: I had drawn the curtains over the window and then said, “please, have a look inside.”

The fix: I deleted that one line. The only thing blocked now is /api/, which nobody outside needs to see.

How to check: Inspect a page with “URL Inspection” in Search Console and look at the screenshot under “Test live URL.” It shows exactly what Google sees. If the page looks broken, Google cannot load its parts.

3. There Were Two Identical Shops on the Street

The whole site opened at www.yotsubagames.com and also at yotsubagames.com without the www. To a person it is the same shop. To a search engine these are two shops at different addresses, and two shops with identical contents leave it unsure which one to list.

The fix: Visitors who arrive without the www are now sent automatically to the www address, with a “we have moved permanently” signal (a 308 redirect).

How to check: Open your site both with and without the www. If both show the site as-is, you have two shops.

4. Japanese Customers Were Being Sent to the English Floor

The site has Japanese and English versions and looks at the browser's language settings to decide which to show. That check was wrong. Visitors whose settings said “Japanese first, English second,” which is completely normal in Japan, were being sent to the English version.

In shop terms: A customer said “Japanese, please,” and was shown to the English floor.

The fix: The check now follows the visitor's order of preference. I also added the language to the links in the header and footer, which removes one unnecessary redirect.

5. “Yes, We Have It” for Things That Did Not Exist

For some addresses that do not exist (/og-image.jpg, for example), the screen said “Page not found” while the server quietly told search engines “all good (200).” The page says “no,” the response says “yes.” Google calls this a “soft 404” and treats it as a sign of low quality.

The fix: Pages that do not exist now answer with a proper “not found” (404). I tidied up old addresses at the same time: pages that moved redirect to their new address, and pages for an old tool I no longer publish answer “gone” (410).

6. A “Grand Reopening Today” Sign, Every Single Day

A sitemap can list a “last updated” date for each page. Mine reported the current time, every time, even for pages that had not changed at all.

In shop terms: A shop that puts up a “Grand Reopening Today” sign every day. People may come by at first, but soon nobody believes it. Google says it only uses the last-updated date when it can verify that the date is accurate.

The fix: The date is now included only for pages whose real update date is known, such as blog posts.

Part 2: Thin or Wrong Content (4 Problems)

7. Shelves Marked “Coming Soon”

The audit found these still sitting in public:

  • A dummy game that had been left in place
  • A “more games coming soon” slot in the card games list
  • Pages for a side project (a 3D printing workshop) with little connection to games. I was focused on game development and had no time to grow it

In shop terms: Empty shelves with a “coming soon” tag. One is fine. A whole row of them makes the shop look like it has not opened yet.

The fix: I deleted the dummy game and the placeholder slot, and moved the workshop pages into storage in a form I can restore later. “Someday” things can go up when someday comes.

8. The Game Rooms Were Dead Ends

Game pages are full screen with no header or footer, so they are comfortable to play. But they had almost no way out to the rest of the site. From Google's side, each one was a dead-end room containing a single game frame.

The fix: Below each game I added a breadcrumb trail, an “Other games” section (six games, favoring ones of the same type), and links to the main pages of the site.

9. The Labels Did Not Match the Products

This one was the most embarrassing. To write the how-to-play guides, I reread the source code of each game, and found that descriptions I had written myself in the game list did not match the actual games.

GameWhat the description saidWhat the game actually does
PyramidThe stock can only be cycled onceIt can be reset twice (three passes in total)
PyramidClear the upper rows firstThe upper rows are covered by the lower ones and cannot be cleared first
Memory MatchMatch the same picture to scoreCards match by number
FreeCell99.99% of games are solvableA number I could not back up

I had gotten the description of my own games wrong. When someone calls that “low value content,” I have no comeback.

The fix: I checked every game description against the code, one by one, and corrected them. Then I wrote nine how-to-play and strategy guides covering rules, controls, scoring, and tips, and added missing explanations to eleven older posts. Where a post had contained a mistake, I said so in the post.

Writing them taught me something: the things only the creator can write are the most original content you have. Numbers like “difficulty rises every 10 points and tops out at 40” or “from the second pass through the stock, each pass costs 100 points” can only come from someone who can read the code. The FreeCell guide, for example, includes the formula for how many cards you can move at once.

10. The House Rules Did Not Match the House

I had not added to the privacy policy for a long time. Meanwhile the site had started creating anonymous accounts automatically, linking Google accounts, publishing rankings and comments, and running analytics. None of it was in the policy.

AdSense requires a site's policy to state that cookies (small records kept in the browser) are used for advertising. If this part is weak, you stumble before your content is even considered.

The fix: I wrote down everything the site actually does and rewrote the privacy policy and terms of service: what is collected, what is shown publicly on the site, which outside services are used, how to ask for deletion, and the rules for posting.

Part 3: Everything Else (2 Problems)

11. My Own Security Guard Was Turning Google Away

Game pages have a safety setting (CSP) that says “only load parts from sources on the approved list.” Think of it as a guard who lets in nobody who is not on the guest list.

To check ad quality, the AdSense script tries to open a small frame inside the page. I had forgotten to put its source (adtrafficquality.google) on the list, and the guard was dutifully stopping it every time.

In shop terms: I applied to show ads, and then my own guard turned away the person who came to check. I do not know whether it affected the review. It is very awkward either way.

The fix: One more line on the guest list.

How to check: Open your browser's developer tools (F12), go to the “Console” tab, and reload the page. A red line starting with Refused to frame ... means something is being blocked.

12. Bonus: Holes That Had Nothing to Do with the Review

Look this closely and you find defects that have nothing to do with the review at all.

  • The posting endpoints took people at their word: The endpoints that accept comments and reviews simply trusted the “I am so-and-so” sent along with the request. They now verify a proof of sign-in (an ID token) on the server and decide who the user is from that.
  • Some sorting options had been broken all along: Choosing “Most popular” for favorites or “Highest rated” for reviews returned an error. The cause was a database index I had forgotten to create in production. I do not even know how long it had been broken.

I only found these because I kept getting rejected, so I am counting that as a win.

Four Things to Check Before You Write Another Article

To everyone stuck in the same rejection loop: before you add one more article, check these four things. Each takes about five minutes.

  1. Compare indexed and not-indexed counts in the Search Console “Pages” report. If more pages are not indexed than indexed, suspect the foundations before the article count.
  2. Use “URL Inspection” to see the page as Google sees it. If it looks broken, check whether robots.txt is blocking the files that make up the page's look.
  3. View the page source and confirm the title and canonical are inside <head>. This is especially worth checking if you use Next.js 15.
  4. Open the site both with and without the www. If both work, redirect one to the other.

For nearly a year I skipped these four checks and stayed convinced that I just needed more articles. I was busily adding pages to a letter that was never being delivered.

So, Did I Pass?

Not yet.

On November 22, 2026, I resubmitted the sitemap and asked Search Console to validate the fixes. Results should take about two to four weeks. I plan to apply again while watching how far those 1,040 not-indexed pages come down.

Pass or fail, I will write the follow-up on this blog. If I pass, it will be a success story. If I fail, it will be a post about hunting for problem number 13. I would rather write the first one.

While we wait, feel free to play a game or two.

🎮 Browse the Games