From an AdSense 'low value content' rejection to a rebuild: what actually changed
In late October one of our content sites received the AdSense "low value content" rejection. It wasn't our first, but this time we resisted the urge to tweak a few articles and resubmit. Instead we treated the entire site as a system to be audited. This post documents the diagnosis and rebuild as a reproducible path for any publisher in the same spot.
First, understand what "low value" is screening for
The rejection letter lists three criteria — provide real information, tools, or services; keep curating content and maintaining structure; attract and sustain user attention. Translated into operational terms, that's three questions: what can a visitor accomplish here? does the content carry unique value? is the site still being maintained?
Sites that get rejected almost always fail the same way: many structurally similar, informationally thin pages that exist everywhere else on the web. A reviewer doesn't need to read everything — sampling ten pages is enough. So step one of self-rescue is not writing new content; it's quantifying how bad things are.
Diagnosis: three metrics
- Visible-text census: strip tags and scripts from every page, count visible characters. The result was sobering — over seventy pages below 1,500 characters, the thinnest at 586.
- Title-pattern analysis: clustering titles by suffix exposed heavy repetition ("guide", "comparison", "pitfalls"), the fingerprint of templated output.
- Publishing-rhythm check: timestamps concentrated in a few days — the signature of programmatic bulk insertion.
Together the three metrics produced a conclusion opposite to our self-image: to a reviewer, the site read as scaled low-quality content.
Four rebuild decisions
One: cut ninety percent of pages. All seventy-plus "tool introduction" pages were deleted. They described other people's products with information available on the vendors' own sites. The sitemap shrank from 160 URLs to about 50 — fewer and real beats many and hollow.
Two: from describing tools to building tools. The new core is eight genuinely usable web tools: word counter, JSON formatter, timestamp converter, password generator, color converter, image compressor, encoder/decoder, and a focus timer. All pure front-end, working instantly on open. "Provide tools or services" stopped being an adjective and became a noun.
Three: from template writing to experience writing. Every retained or rewritten article comes from a project we actually shipped: payment migration debugging, mini-program rejection retrospectives, build-chain pitfalls. Structures differ — retrospectives, checklists, tutorials — every piece above 1,500 characters, publish dates spread across a realistic editorial cadence.
Four: trust pages a human can verify. The About page carries the company's registered name and background; Contact lists a real mailbox with a response-time commitment; Privacy explains exactly what is and isn't collected. Nobody reads these pages daily, but reviewers always do.
Don't let the tech side add friction
Three hard defects fixed along the way: the site loaded adsbygoogle.js but lacked the <meta name="google-adsense-account"> tag — both must be present; the sitemap was regenerated and every URL checked live; robots, 404 handling, and mobile responsiveness audited. None of these alone cause rejection, but together they raise review friction.
Results, and an honest caveat
Two weeks after relaunch we resubmitted. The changes were quantifiable: thin pages from seventy-plus to zero, average visible text up fourfold, real interaction data on tool pages. The review passed — but the more important gain came after: organic search traffic began a slow climb, because every remaining page now deserved indexing.
To be honest about the caveat: rewriting old articles into a new template would most likely have failed again. The opposite of low value is not a prettier template — it's the things only you could have written. Shipped projects, debugging war stories, hand-built tools: that's the value both reviewers and users recognize.