How Thematic uses AI-powered discovery to personalize music recommendations at scale

Ready to improve your music discovery workflows? Try Similarity Search in Cyanite.

Thematic is a creative community built for discovery, collaboration, and growth. Co-founded by Michelle Phan, a pioneering creator who helped define how influencers build and monetize audiences online, Thematic was optimized for the creative experience from the start and is now trusted by over 1 million creators.

When a creator features a Thematic song in their video, they create a promotional moment for that artist. The artist can track exactly which videos by which creators used their music and how many new fans they gained as a result.

Every interaction feeds back into the recommendation engine, creating a virtuous loop of value between creators and artists. It’s a collaborative ecosystem where both sides of the creative equation create, connect, and grow together.

Thematic’s goal was to turn music discovery into a win for all by giving creators better options. However, in practice, their growing catalog meant creators had to navigate an overwhelming amount of choice in a space already defined by decision fatigue. As Thematic continued to scale, so did the pressure to make discovery smarter.

When finding the right track becomes a problem

Before building anything new, the team listened carefully. One-on-one interviews, surveys, support ticket analysis, and community feedback all pointed to the same problem. Finding the right song was taking too long.

The size of the catalog wasn’t the issue. It was the mismatch between how creators think and how search tools work.

Tags and genres are a poor way to describe music. “Creators don’t fall in love with a song because it’s ‘indie folk with a male vocalist at 120 BPM,’” explains Audrey. “They connect with it because of how it sounds, how it makes them feel, and the emotional resonance of the lyrics.”

Traditional keyword search asks users to “translate a feeling into a music tag formula,” as Audrey puts it. Most creators know what they want when they hear it. Forcing people into a keyword search box creates a time-consuming cycle: sampling track after track, exploring a genre only to find it missed the mark, then starting over.

But Thematic serves both creators and artists, so improving discovery efficiency would have a double impact: saving creators time finding the perfect-fit song, while driving higher placement opportunities for artists.

“Think about the difference between flipping through cable channels versus opening a feed that already knows what you like,” says Audrey. “The channel count doesn’t matter. What matters is whether the right thing surfaces at the right moment.”

To address this, Thematic launched a rebuilt platform alongside a complete rebrand. The product had grown significantly since its initial launch, and the visual identity needed to catch up.

The rebuild focused on two areas: 

• A personalized For You experience based on each creator’s usage history and music taste, updated tagging infrastructure to improve search and filtering accuracy, and the ability to find sonically similar songs to any track, whether on Thematic or Spotify.
• Deepening the creator community through a points leaderboard, upgraded creator and artist profiles, and better visibility into the value exchange happening on the platform.

At the center of the smarter discovery experience was Cyanite.

The decision to incorporate AI-based audio analysis came down to one thing: saving creators time.

“We treat AI as a great workflow improvement tool for creators,” says Audrey.
“Its ability to analyze large datasets and surface the most relevant information can be genuinely time-saving, especially when trying to identify songs that have a similar sound, not just similar music attributes and tags.”

Sound-based search has removed the need to translate intent into search terms. Instead of asking creators to describe what they’re looking for, it lets them start from a reference track and find cleared, licensed music on Thematic that matches that sonic profile. What used to take hours can now happen in seconds.

From there, creators can quickly build a full set of songs that fit their overall aesthetic by exploring complementary recommendations. What used to be a slow, trial-and-error process becomes fast, flexible, and far more creatively aligned.

Why Cyanite?

When evaluating solutions, Thematic made their decision based on capability and practicality.

Beyond features, the path to implementation mattered too: “Product-switching a core backend component can often be painful, but the ease of transition was seamless.”

How Cyanite fits into Thematic’s workflow

Cyanite is integrated across several touchpoints, allowing Thematic users to move fluidly between different discovery pathways depending on how they work.

Auto-Tagging

Auto-Tagging is the foundation of everything. It provides the baseline context for every song in Thematic’s database and powers the filtering, matching, and personalization that follows. For artists who haven’t provided their own descriptions, Cyanite also generates auto-descriptions, ensuring no track enters the catalog without proper context.

This is important for emerging artists. “Discoverability can often be a visibility problem dressed up as a quality problem,” says Audrey. “A great song that isn’t tagged correctly, or that gets added to a platform when the algorithm isn’t paying attention, can sit undiscovered for months.” Accurate, consistent tagging prevents that gap from opening up.

Similarity Search

Thematic’s Similarity Search lets creators find music that matches a reference track’s sonic and emotional profile. For any song, one click surfaces a playlist of sonically similar tracks. It also works for any external song via Spotify URL: entering a link returns a curated shortlist of cleared music that matches the sound.

The feature directly addresses the feedback creators gave Thematic in their research. No more guessing at genre labels or struggling to articulate the vibe they’re after. The reference track does the talking.

Advanced Search

Advanced Search powers part of Thematic’s For You page by recommending songs based on what a specific creator has already downloaded and used.

“Advanced Search might be my favorite” says Audrey, “as it’s a hyper-personalized way for us to provide perfect-fit song recommendations for our users, contextualized to songs they already like to use. It’s not just, ‘Here are songs in this genre.’ It’s ‘Here are the songs most likely to work for you, based on everything we know about how you create.’”

Cyanite’s multi-track similarity underpins this. Rather than building a recommendation from a single song, Thematic builds a rich, dynamic music profile for each creator from their full history. The result is recommendations that feel more accurate than single-song matching can achieve on its own.

The new Thematic homepage is the clearest expression of what this infrastructure makes possible.

Every module is personalized. Depending on a creator’s behavior and preferences, they might see new drops from artists and creators they follow, song suggestions based on their download history and content niche, editorial playlists matched to trending formats, and exclusive creator perks and rewards.

One of the modules is powered directly by Cyanite’s Advanced Search. Others are shaped by the music tags Auto-Tagging produces. Together, they create a feed designed to surface the right song before the creator even has to go looking for it.

The impact of integrating Cyanite

Since integrating Cyanite, Thematic has seen consistent, meaningful changes in how creators search and make decisions:

• Nearly 9% decrease in the number of plays before a creator downloads a song
• Nearly 15% increase in creators downloading songs directly from the For You home page, without using search or filters at all

Creators are reaching confident decisions faster, spending less of their energy on searching and editing and more on actually creating.

The second signal is the one that excites Thematic most. “We’re getting the right song in front of the right creator before they even have to go looking for it,” says Audrey. “That’s the goal: discovery that feels effortless.”

For artists, the impact is equally direct. Better matching means more placements, more exposure, and more new fans. AI-powered similarity matching means a song’s sonic profile does the work that metadata alone can’t. As Audrey puts it, “The right creator finds the right artist not because they searched for them, but because the system recognized the match.”

Rethinking where music fits in the creative process

Beyond the metrics, Thematic sees a larger shift ahead. Music has traditionally been the last step in a creator’s process. They tend to hunt for the perfect fit after the rough cut is done and the overall vibe is already set. AI-powered discovery has the potential to flip that entirely.

When Thematic decided to invest in AI-powered discovery, removing friction was the priority: for creators looking for the perfect track, and for artists waiting to be found. Cyanite made both possible at once.

Improve your catalog’s discoverability with Cyanite. Sign up to get started.