How Ariadne Decides What’s Related

The short version

Ariadne looks at four things about your content and combines them into one score per candidate post:

• Shared tags — how many tags two posts have in common.
• Shared categories — the same idea, for categories.
• Title similarity — how much meaningful overlap there is between two titles’ words.
• Body similarity — how alike the actual writing is, both in shared wording and (with AI matching enabled) in meaning.

Each signal is scored independently, then combined using weights you can adjust from Settings → Matching. By default, tag overlap counts for the most, followed by body similarity, category overlap, and title overlap.

Why four signals instead of one

Any single signal on its own has an obvious failure mode. Category-only matching is often too broad. Title-only matching misses posts that are genuinely related but titled differently. Ariadne combines multiple signals specifically so that a post has to be relevant in more than one way to rank highly — and because every signal is visible and scored, a match is never a mystery.

Shared tags and categories

This is a simple overlap count, capped after five shared tags (or categories) so a post that happens to share dozens of tags with another doesn’t dominate the ranking disproportionately — five shared tags scores the same as fifteen.

Title similarity

Ariadne strips common filler words (like “the,” “and,” “of”) from both titles and compares what’s left. This is a genuinely separate signal from body similarity below — two posts can share almost no title words but still be strongly related in their actual content, or vice versa.

Body similarity

This is the most substantial signal, and it works in two layers:

• Local text matching runs entirely on your own server, comparing the actual words used in each post’s title and content. This works for every site, with no external service and no ongoing cost.
• AI-powered semantic matching (Pro and Pro+) additionally understands meaning, not just shared words — two posts about the same underlying topic, written with completely different vocabulary, can still score as closely related. See do you need an AI API key?

When both are available, they’re combined so that a post ranking well on either count contributes to the final score — a post can be a strong match through exact wording, through meaning, or both.

Adjusting the weights

If your site has an unusual content shape — heavily tagged posts, very short titles, or very long-form articles — the default balance might not fit. Every weight (tag, category, title, body) is a plain slider on the Settings screen. Changing a weight recalculates every post’s related list in the background; nothing needs to be re-saved manually.

For the curious: what “candidates” even means

Before scoring happens at all, Ariadne narrows down which posts are even considered — by default, only posts sharing a category. This candidate pool is what gets scored on the four signals above, and it’s configurable under Settings → Matching → Candidate pool. See where related posts show up by default for the full default configuration.