What Google means by low value content

In August we applied to run ads on this site and were turned down. The reason given was low value content, with four policy documents linked and no specifics about which parts were the problem.

That vagueness is the common complaint, and it is worth writing up honestly, because we have since watched several clients receive the same message and read it as arbitrary. It isn't. It just describes an outcome rather than a cause.

Measure the site the way the reviewer does

The first useful step is refusing to argue about it and counting instead. Strip the markup from every page and count the words that remain. That is roughly what an automated pass sees.

curl -sL https://example.com/page | python3 -c "
import sys,re,html
h=sys.stdin.read()
h=re.sub(r'(?is)<(script|style)[^>]*>.*?</\1>',' ',h)
t=html.unescape(re.sub(r'(?s)<[^>]+>',' ',h))
print(len([w for w in t.split() if any(c.isalpha() for c in w)]))"

Our result: about 2,000 words across eight pages, averaging 254 per page, with no articles, no privacy policy and no terms. Written out like that, the rejection stops looking arbitrary. It was a brochure asking to be treated as a publication.

What the phrase actually covers

Reading the four linked documents together, low value content is doing the work of several distinct problems:

• Not enough substantive text for a reviewer to evaluate. A site of navigation and headlines has nothing to assess.
• Nothing original. Content that restates what is already widely available adds no reason for the page to exist.
• Missing the pages a legitimate publisher has. A privacy policy is effectively mandatory, and its absence alone can end an application.
• Content produced at scale to fill space rather than to be read, which has its own harsher policy category.

That last one matters, because the instinctive fix is to generate thirty articles quickly. Doing so trades a soft rejection for a spam classification. Scaled content abuse is judged on intent and quality, not on whether a human or a machine typed it.

What we actually did

We wrote eleven long-form pieces, each about a specific system we had built and the way it had failed. A deploy pipeline standing in for a backup. A metric that flattered its own results. A ranked list where the leader turned out to be one small source counted repeatedly.

The specificity is the point. Anyone can write about the importance of backups. Only someone who ran the audit can tell you that three of four repositories had no remote while every deployment was green, and that is the sentence a reviewer cannot find anywhere else.

Alongside that: a privacy policy disclosing advertising cookies with working opt-out links, terms, and a rewrite of the two thinnest sales pages. The site went from about 2,000 words to about 13,500, and from eight pages to twenty-two.

Two details worth stealing

Put a name on the writing. Articles attributed to an organisation give a reviewer nothing to weigh. A named author with a role, a short biography and a link to a real professional profile is a materially different signal, and it costs an hour to add.

Make the privacy policy match the site's actual behaviour. If a consent banner asks permission to set advertising cookies while the policy never mentions advertising, the two documents contradict each other, and one of them is wrong in a way that is easy to check.

The honest part

At the time of writing the review is still pending. We can tell you what the rejection cited and what we changed. We cannot tell you it worked, and anybody who promises a specific outcome from a subjective review is selling something.

What we can say is that the exercise was worth doing independently of the result. The site is better, the sales pages answer questions they previously ducked, and the writing has brought in conversations that ads never would have.