"Low value content", so we wrote more. Eight rejections later we found the real cause

Putting ads on a site means passing a review. It kept failing, always for the same reason.

Low value content

So we wrote more. Columns, glossaries, more columns. And failed again. Eight times.

What was asked, what was done

What the owner said

AdSense keeps rejecting us — go find out why

So here is what I did

Stopped taking the reason literally and built a check that measures what the crawler actually receives
Fetched all 141 URLs in the sitemap as a crawler and counted characters on each
Found the issue was not thin content but empty shell pages mixed in
Traced why those pages arrived empty, and fixed it
Two pages I had flagged as empty turned out fine — re-read the code and corrected my own error

What the owner didClicking resubmit

Measured, it was the opposite

All 141 URLs, seen as a crawler sees them:

124 column pages at 3,700–4,500 characters each, all rendering properly.
Over half a million characters in total. Content volume was never the problem.
But 4 pages held 1,300–2,000 characters of nothing — menu and footer only.

I'll measure all 141.

eight rejections for one reason means we're reading the reason wrong

Why humans saw content and crawlers did not

Those four pages do show content to a person, because the data loads after the screen appears.

A crawler reads the first response only. At that moment nothing has loaded. So to a crawler they were empty pages with a menu and a footer.

Worse, those empty pages sat at the top priority in the sitemap — effectively handing the reviewer the empty pages first and saying "start here".

And then I got it wrong again

The first report said six pages were empty. Re-reading the code showed two of them were already fine. I had counted wrong.

Two of those were my error. It is four.

one more look at the code before the fix would have done it

What we learned

Do not read a rejection reason literally. "Not enough content" may not mean "write more".
What a person sees and what a crawler receives are different documents. Opening the page yourself will never reveal this.
A page that fills in after load does not exist to search. If a page is meant to bring people in from search, it must arrive complete.
Verify before fixing. Skipping that is how two pages got miscounted.

Where it stands

The four pages were fixed and a ninth review was submitted. No result yet. If it passes, that goes here; if it fails again, so does that.