Thin against your own site: outliers by page type

What a site-relative thin page is

An absolute word-count threshold is wrong on most sites. Product pages, documentation pages and long-form articles all have different natural lengths, and one number cannot judge them together.

Your own site already sets the expectation. If your articles median 1800 words, a 300-word article is an outlier against its siblings whatever an absolute floor says. This rule groups pages by type and reports only the low outliers within a group.

What squirrel checks

The rule is site-wide. It groups crawled pages by page type, then computes the median and the median absolute deviation of the word count in each group. It emits a single check named thin-vs-site-norm:

• Skipped when there are no pages. Message: No pages available for analysis.
• Skipped below the site floor of 10 crawled pages. Message: Only 6 page(s) crawled; 10 needed to establish a site norm.
• Skipped when no page type qualifies. Message: No page type has 8+ comparable pages with a clear word-count norm.
• Pass when no page is an outlier. Message: All 210 comparable page(s) sit within their page type's word-count norm (median words: article 1840, product 320).
• Warn when outliers exist, and fail when they cover 20% or more of the judged pages. Message: 7 of 210 page(s) are far shorter than the norm for their page type (median words: article 1840). Each item reads 340 words vs the article norm of 1840 (18%), with up to ten listed.

Four guards apply in order. The site needs 10 crawled pages. A page type needs 8. At least 70% of a group must sit within half its median for that median to count as a norm. And a flagged page must be both under half the group median and more than three deviations below it, where a deviation is the larger of 1.4826 times the median absolute deviation and a tenth of the median, so neither a tight nor a wide distribution can fire on its own.

Only pages that returned a 2xx status and carry a known page type are sampled.

Pages below 300 words are never reported here, because content/word-count owns them, but they do still count toward their group’s size, median, deviation and agreement, and toward the judged-page total in the message. Severity warning, weight 3.

How to fix it

Expand the page to the depth its siblings have, consolidate it into a fuller page, or noindex it if it exists only for navigation.

A deliberate stub or a landing page that is short on purpose will be reported. The finding is that it does not match what the rest of its page type promises, which is worth confirming rather than padding.

Enable / disable

Disable this rule

[rules]
disable = ["content/thin-vs-site-norm"]

Disable all Content rules

[rules]
disable = ["content/*"]

Enable only this rule

[rules]
enable = ["content/thin-vs-site-norm"]

enable replaces the default ["*"], so this is the only rule selected. Leave disable out: a rule that matches any disable pattern is skipped even when enable names it. Only an enabled setting under a rule’s own [rule_options] table overrides both lists.

Related rules

• content/word-count: the absolute floor this rule defers to.
• content/title-pattern-outlier: the same outlier approach applied to titles.
• url/slug-convention: the same approach applied to URL conventions.
• content/duplicate-title: pages that are similar in a different way.

Content findings ship in every audit next to the SEO, performance and agent experience rules. See Fix SEO issues with AI for how an agent works through a report.

References

• Creating helpful, reliable content, Google Search Central

Check your site

Run squirrel audit https://example.com and open the Content section of the report. Each outlier is listed with its word count and its page type’s median. Signed-out local audits need no account and cost nothing. Signed-in audits spend credits.