Back to sign inPairMuse

Legal

Transparency: ranking and AI

Version 2026-10Effective from October 18, 2026

German is the reference version of our notices and information. Your contract is governed by the contract language you chose when you ordered.

This page explains how PairMuse orders posts, products and suggestions, and how we mark text generated by AI.

01 Home: Following view

The Following view on Home shows posts from the people you follow by time of publication, newest first. Here we do not rank by your behavior. Posts from people you have restricted do not appear. Filters narrow the view, for example by color, material or brand.

If you choose a sort such as Popular, Most liked or Most commented, we order by that value. Popular counts each like once and each comment and each save twice.

02 Home: Discover view

The Discover view shows your own published posts at the top, newest first, and below them posts from everyone you are allowed to see, ranked for you. The candidates are posts our search finds among them, posts that are popular right now and posts from people you follow or people they follow. The ranking uses:

  • Your connection to the person who published the post (weight 1). It grows when you like, comment on or save their posts: a comment counts one and a half times as much as a like, a save twice as much. “Dislike” lowers it as much as a like raises it, a report three times as much.
  • How new the post is (weight 1). Its value halves with every day.
  • How popular the post is (weight 0.5): likes, comments and saves, dampened so that popular posts do not crowd out new ones.
  • How well our search rates the post (weight 0.5): how closely its text matches your search term, if you entered one, and how new it is. This does not reflect your taste: we derive no taste from the categories or features of the posts you interact with.
  • How often we have already shown you the post. Each time its value drops by 35 percent; posts you have seen often are hidden.

We then mix the order so that no single person or style fills the page. This ranking applies as long as you do not choose another sort. Filters narrow the view; a sort such as Popular replaces the ranking for you. If you want posts without any ranking by your behavior, the Following view shows them in time order.

04 Suggestions from Muse and Stylist

Muse and Stylist look for pieces for you first in your wardrobe, then in the PairMuse catalog, and only then, if needed, with a live search of the shops you selected in Settings.

Conditions come first. A piece qualifies only if it fits your request, such as the category or color you asked for, and your saved preferences: department, fit, style direction, occasions, allowed materials and price ceiling. If you expressly say something different in the conversation, that applies instead of the saved preference. If a detail of a product is unknown, we do not rule the product out for that reason; the assistant’s AI model then judges whether it fits.

Among the pieces that qualify, what counts first is whether a piece is exactly the category you asked for, and then the judgment of a re-ranker, an AI model on our own server that rates how well the piece fits your request. From these results the assistant’s AI model picks the cards you see. We receive no money from shops for the selection or the order.

05 No paid placement

No shop and no brand pays for products or posts to appear on PairMuse or to rank higher. If you buy from a shop through a link, we receive no commission; we use no affiliate links. If that changes, we will clearly label paid content.

Comments on products come from members with an account. We do not check whether they bought or wore the product.

06 How AI text is marked

Muse and Stylist are AI assistants. A line under the input field of every conversation with Muse or Stylist says that the assistant is an AI and can make mistakes.

Their replies are produced by models from OpenAI, Anthropic, xAI and Google. Where the provider supports it, these models embed an invisible, machine-readable watermark in the text.

Not marked as AI text are product cards that only select and order existing products, short labels and values such as colors or sizes, and the details an AI reads from the photos of your pieces.