Plex recommendation tools compared

There are a lot of these now, they overlap confusingly, and the READMEs all say “personalized recommendations”. This page is an attempt at a straight comparison of what each one actually does.

Disclosure: Shortlist is my project, so read the row about it with that in mind. I’ve tried to be accurate about everything else and to say plainly where another tool is the better pick. Facts checked 14 August 2026 against each project’s repository — this space moves fast, so verify anything that matters to you. Corrections welcome.

Sent here after trying Kometa or SuggestArr?

If you asked for “a tool that gives each Plex user recommendations based on their watch history”, the three names you were almost certainly given first are Kometa, SuggestArr and Recommendarr. They’re the biggest and best-known projects near this problem, and it’s worth being precise about why none of them answer that particular question — because working it out by installing all three takes an evening.

Kometa (3,400+ stars, formerly Plex Meta Manager) is a metadata and collection builder. It creates collections from rules you write in YAML — by genre, by decade, by Trakt list, by almost anything — and manages artwork and metadata beautifully. It is the right tool for shaping a library, and it’s excellent at it.

It is not a recommender. It doesn’t read anyone’s watch history to decide what to suggest, and its collections are library-wide: everyone with access sees the same ones. Shortlist is built to run alongside Kometa rather than instead of it — it never touches a collection it didn’t create.

SuggestArr (1,200+ stars) watches what you recently played and automatically requests similar content through Jellyseerr/Overseerr, so your library keeps growing with things you’ll probably like. Works with Plex, Jellyfin and Emby.

It’s about acquisition, not presentation. Its output is new files on your disk, not a row on anyone’s home screen — and the “recently watched” trigger is server-wide rather than a per-person profile. If your complaint is “my library never has anything new”, it’s a great fit. If it’s “my library is huge and nobody can find anything in it”, it’s solving the opposite problem.

Recommendarr is AI-driven and works from Radarr/Sonarr library data. Be careful searching for it — there are several unrelated repos under that name and near-spellings, so check what you’re installing. Like SuggestArr it’s oriented toward what to add, not toward per-user rows in Plex.

The common thread: all three are server-wide. None of them produces a different row for different people, and none of them hides one person’s row from everyone else. That’s not a flaw — it’s just a different problem, and it’s the problem the rest of this page is about.

First, decide which problem you have

The tools split into four groups, and picking from the wrong group is the usual mistake.

“Keep my library stocked with things we’ll like.” You want new content acquired automatically. → SuggestArr, Recommendarr.

“Organise the library I already have.” You want well-built collections, artwork and metadata. → Kometa.

“Help me decide what to watch tonight.” You want a dashboard you open when you can’t choose. It doesn’t need to touch your Plex library at all. → SeekAndWatch.

“Give each person on my server their own rows.” You share with family or friends and want each of them to get suggestions from their own viewing — ideally without everyone else seeing them. → the rest of this page.

That last one is where the privacy question appears, and it’s the axis the remaining tools differ on most.

The comparison

Kometa and SuggestArr are in the table because people arrive having tried them, not because they compete on per-user rows — see above for what each is genuinely for.

Shortlist

Per-user “Picked for You” rows built from each person’s own watch history, made private with Plex’s label restrictions — every other account’s share filter gets label!=shortlist_<user> merged into it, so a row is visible only to its owner. Rows are delivered unpromoted, exclusions merged, and only then promoted, so a row is never visible before the rule hiding it exists.

Share filters are snapshotted before the first write and restored exactly on uninstall; it merges rather than rebuilds them, skips the owner, and never modifies a collection it didn’t create. Everything supports --dry-run. AI is optional — the built-in picker needs no keys.

Plex-only, and it will stay that way: the privacy model depends on Plex’s label-based share filters, which Jellyfin and Emby don’t have. Needs Plex Media Server 1.43.2.10687+ and a Plex Pass on the admin account. MIT.

Immaculaterr

The most feature-dense of these. Event-driven rather than purely scheduled — it reacts when someone finishes something and can build rows immediately, alongside off-peak refresh, discovery and library cleanup. Builds a lot of named collections (“Based on your recently watched”, “Change of Taste”, “Fresh Out Of The Oven”), supports profiles with their own users, media types and filters, and integrates with Radarr/Sonarr.

Gives each monitored viewer separate rows and separate history, and pins rows to surfaces that viewer can see — but it doesn’t claim per-user privacy, which is a different thing from per-user content. Ships on both GHCR and Docker Hub. Licensed under custom terms rather than a standard OSI licence, which is worth reading if that matters to you.

Pick this if you want the richest set of automatic collections and a tool that reacts in real time, and you’re relaxed about other users seeing each other’s rows.

Curatarr

Analyses each user’s watch history and scores unwatched library content by keyword, genre, cast and director similarity, creating per-user collections that update automatically. Also generates external watchlists so you know what to acquire next, and integrates streaming-service availability.

Its per-user separation is explicit in its own documentation about being a UI-level split rather than access control — users’ collections are separated in Browse and Search, not hidden from each other by Plex permissions. That’s a reasonable design choice, clearly disclosed, and fine on a server where everyone’s relaxed about it.

Distributed as self-updating signed binaries for Windows, macOS (Apple Silicon) and Linux as well as from source — no Docker required, which is genuinely the easiest install here. MIT.

Pick this if you want per-user recommendations without running Docker, care about what to acquire next as much as what to watch, and don’t need rows hidden from each other.

SeekAndWatch

A different category: a “what should we watch?” dashboard rather than a row builder. Connects Plex, Tautulli, TMDB, Radarr and Sonarr in one place, with Smart Discovery from your watch history, a Kometa config builder that saves you writing YAML, Overseerr requests and Tautulli trending. There’s a hosted Cloud beta so friends can request without access to your apps.

Pick this if the real problem is you and your household staring at the library unable to choose, or if you want a Kometa config builder — that feature has little competition.

TV-Show-Recommendations-for-Plex

A well-documented Python script rather than a service. Builds a taste profile from watch history — genres, cast, crew, keywords, weighted by your Plex ratings — scores unwatched shows, and can label them in Plex or push new titles to Sonarr via Trakt. Reads other users’ history through Tautulli, and runs attended (confirm each pick) or unattended. A companion Movie Recommendations script covers films.

Last updated March 2025, and no licence file — which by default means all rights reserved, worth knowing before you build on it.

Pick this if you want something small you can read end to end and drive from cron, and you like approving picks by hand.

plex-recommendations-ai

Creates a single “Recommended” collection using OpenAI over your watch history, with a generated summary explaining the choices. Simple, cheap to run, and one of the earliest tools in this space.

Not per-user, and unmaintained since May 2023 — listed here because it still ranks well in search results and people find it first. No licence file.

The three questions that actually narrow the list

1. Do you want content acquired, or content surfaced?

The most common wasted evening in this space is installing an acquisition tool to solve a presentation problem. SuggestArr and Recommendarr add files to your library. Kometa organises the files you have. Neither changes what any individual sees on their home screen. If your library is already big and the complaint is “nobody can find anything”, you want the surfacing group.

2. Do you need rows to be private, or just personal?

Those get conflated constantly and the difference is the whole architecture. Personal means the titles are chosen for one person. Private means nobody else can see the row. Most tools do the first. Doing the second requires Plex’s label restrictions on share filters, a minimum server version, a Plex Pass, and careful write ordering — which is why most tools don’t.

On a two-person household, personal is plenty and you should pick on features. On a server shared with fifteen friends, twelve rows called “Picked for Dave” on your home screen is a clutter problem before it’s a privacy one — and a row built from someone’s viewing habits is more revealing than people expect.

3. Does it need to coexist with what you already run?

Most people asking this question already have Kometa, Radarr and Sonarr in place. A recommender that rewrites collections it didn’t create, or that fights another tool over shelf order, costs more than it adds. Check that whatever you pick scopes its writes to its own collections — and if you run Kometa, that the two won’t overwrite each other.

Answer those three and the list narrows to two or three.

Related

• How to improve Plex recommendations — before installing anything
• AI recommendations for Plex — which tools need an API key and why
• Per-user collections — the privacy mechanism, and how to do it by hand
• Recommendations from watch history — what to look for in any approach
• Getting started — installing Shortlist
• FAQ — Kometa coexistence, what the owner can see, uninstalling