# Pages far from the site's subject

> Which pages are about something else. Every embedded page gets a Site fit — how close its meaning is to the site as a whole — and the pages unusually far from the rest are listed under one filter. A stray post, injected spam or a section that needs its own home shows up here.

Describes GLLmolfar 0.9.35, checked 2026-11-08. Canonical page: https://www.getlooploop.com/gllmolfar/docs/off-topic

- **Where:** Site › All pages › More columns… › Content › Far from the site's subject
- **Mode:** Pro
- **Built from:** The embeddings of the open crawl's pages
- **Changes:** When the crawl's analysis ends

## On the screen

The Content column set on the filter **Far from the site's subject**, with a page picked and its **Similar pages** open below.

![GLLmolfar's Content tab filtered to Far from the site's subject on a crawl of getlooploop.com: 22 of 500 pages, with the Similar pages of /es/glossary/anova below](https://www.getlooploop.com/images/gllmolfar/docs/off-topic.webp)

Numbered parts of the screenshot:

1. **Far from the site's subject.** The filter: 22 pages, 4.4% of the crawl, fall below this crawl's cut.
2. **The pages.** Each page far from the site's average meaning. Here, glossary and notes pages in other languages and the cookies policy.
3. **Closest page by text, Text similarity.** The Content columns for near duplicates by text: empty here, as none of these pages has a close twin by words.
4. **Similar pages.** The pages nearest in meaning to the picked one, with their similarity: whether it has neighbours on the site at all.
5. **Similarity.** From 0 to 1. Its own translations come first; the rest of the site sits much lower.

## How far is far

- **Site fit** is the cosine between a page's embedding and the mean of all the crawl's page embeddings: what the site as a whole is about. Add it with **Columns** to any page column set.
- The cut is the median Site fit less three median absolute deviations. A site whose pages all share one subject flags none; a page about something else stands out however many pages there are.
- With fewer than 20 embedded pages there is no cut and the filter stays empty: too few pages to say what is usual.
- Pages that were not embedded (embeddings off, or pages the embeddings skip, such as page 2 of a list) have no Site fit and are never listed.

> **An insight, not an error.** The fix list rates it as an insight. A page can be far from the rest on purpose — a legal page, a careers page. Open the pages: keep what belongs and link it from a related section, and remove or move what does not.

## Related pages

- [Topics](https://www.getlooploop.com/gllmolfar/docs/topics.md): The Topics tab: the site's pages grouped by meaning, with each topic's size, key terms, cohesion, internal linking, cannibal pairs and central page.
- [Topic map](https://www.getlooploop.com/gllmolfar/docs/topic-map.md): The Topic map tab: every embedded page as a dot, placed by meaning and coloured by topic, status, indexability or depth, with a card for each page.
- [Duplicates and canonicals](https://www.getlooploop.com/gllmolfar/docs/duplicates-canonicals.md): Exact duplicates, near duplicates by text and by meaning in the Content column set, and the Canonicals column set with its filters and report.
- [Settings › Embeddings](https://www.getlooploop.com/gllmolfar/docs/settings-embeddings.md): Settings › Embeddings: whether new crawls build embeddings, how much of each page the model reads, and whether it runs on the CPU or the GPU.
- [Fix list](https://www.getlooploop.com/gllmolfar/docs/fix-list.md): The Fix list tab: what to fix in the crawled site, numbered, most important first, each with why it matters, how to fix it and the URLs.