Using MyJay Search.

MyJay Search indexes public pages from MyJay, Neocities, and Nekoweb. It's a plain ranked match on what you typed, not a tuned-for-engagement algorithm. For how indexing itself works (and how to opt out), see search indexing.

The basics

Type a query and hit Search, or pick a suggestion as you type (autocomplete is matched against terms already in the index, so a suggestion always returns something). You can type a single word, a few keywords, or a full sentence, all three work.

How results are ranked

Similarity comes before relevance. Every result is scored on two separate things, in this order:

  1. Similarity: how much of what you typed a page actually covers. A page matching 4 of the 5 distinct words in your search outranks a page matching only 1 of them, full stop, regardless of anything else.
  2. Relevance: only used to break ties between pages tied on similarity. This is where word position (title beats a meta description, which beats the body text), word rarity (a distinctive word counts for more than a common one, the same idea behind TF-IDF if you've heard that term), and an exact phrase match (see below) all come in.

This is deliberately different from requiring every word to match. A search box that demands all of your words appear on the same page falls apart the moment you type more than two or three words, since real pages almost never contain a full sentence verbatim. Ranking by coverage instead of filtering on it means a long, sentence-length query degrades gracefully: the closest match is still first, and you never get a blank page just because you typed in full sentences instead of keywords.

Exact phrases

Wrap part of your query in quotes, "like this", and a page containing that exact phrase (anywhere in its title, description, or body) gets ranked well above one that only matches the same words scattered separately. It's a bonus, not a requirement: a page that doesn't contain the literal phrase can still show up ranked by ordinary word matching, quoting something narrows the ranking, it never narrows you down to zero results.

Filters

The sidebar (or the filter section above results on narrow screens) lets you narrow by:

Filters and your query are part of the URL (/search?q=...&platform=...), so a search result page is always a link you can share or bookmark.

"Did you mean"

A single-word query that returns nothing gets checked against terms already in the index for anything close (a short edit distance, same first letter and similar length). If something close exists, it's offered as a suggestion, click it to re-run the search.

No results, never a blank page

An empty result set always offers something else to do: browse recently indexed sites, a tag, or hit "a random site." The homepage of /search (no query) shows the same recent feed and tag list, plus a Surprise me button that opens one random indexed page in a new tab.

Sites like this

Every result card has a "Sites like this" link that expands in place, showing other indexed pages that share tags with it (or, for an untagged page, other pages on the same platform). Nothing is fetched until you click it.

What this isn't tracking

Searches are logged as plain query text (for the admin's "what are people searching for" view, and to see which queries return nothing), with no IP address, no account link, and no record of which result you clicked. Consistent with the rest of the platform: no trackers, no algorithm tuned on your behavior.