What 72,000 AI Songs Reveal About How People Actually Use AI Music

We analyzed 72,637 AI songs that 21,794 people made on MusicWave between October 2025 and October 2026. Love songs come first, worship songs second and family songs third. About 1 in 8 songs is made for a specific person, and 1 in 5 is prompted in a script other than the Latin alphabet.

Every chart and number on this page is free to reuse. All we ask is a link back here.

The key findings

• Love is the #1 theme. 14.8% of the songs people described in their own words are about love or romance.
• Faith and worship is #2 at 8.1%, ahead of family (7.2%), parties (4.4%) and friendship (3.6%).
• About 1 in 8 songs (12.7%) is made for someone: a partner, a parent, a child, a friend, a birthday, a wedding or a memorial.
• 1 in 5 songs (20.6%) isn't prompted in the Latin alphabet. Arabic script alone accounts for 12.3%.
• Song-making grew 11x in seven months, from 2,431 songs in March 2026 to 26,664 in October.
• More people bring their own lyrics. 37% of October's songs started from a full set of lyrics, up from about 1 in 4 over the summer.
• The average AI song runs 3 minutes 22 seconds, and 1 in 4 runs past four minutes.

What do people write songs about?

The stereotype says AI music is background filler for videos. The prompts tell a much more personal story.

Love and romance lead at 14.8%. Think songs for a partner, proposals set to music and anniversary songs with a name in the chorus.

Faith and worship comes second at 8.1%, covering worship songs, hymns, gospel and Islamic nasheeds. It's a category that barely comes up when people talk about AI music.

Family comes third at 7.2%: songs for mom, dad, a sister, a son or a grandmother. If you're looking for ideas there, our lists of songs for your sister and songs for dad are a good place to start.

After that comes a long tail of everyday life. Parties, friends, birthdays, school, pets, money and the holidays all show up.

A song can land in more than one theme (a birthday song for mom counts as both), so the shares add up to more than 100%.

1 in 8 songs is made for someone

Count every song that names a partner, a family member or a friend, or that marks a birthday, a wedding or a death. You get 5,825 songs, which is 12.7% of everything people described in their own words, or about 1 in 8.

Across the whole library, we counted 1,280 birthday songs, 611 songs for weddings and proposals, and 477 songs made in memory of someone who died.

That last number is the one we keep coming back to. People don't make memorial songs to chase streams.

They play them once at a service, or send them to a brother or sister who's grieving. For a lot of people, an AI song is closer to a card or a letter than to a product.

1 in 5 songs isn't written in the Latin alphabet

We checked the writing system in every song's prompt and title. 20.6% use a script other than the Latin alphabet.

Arabic script is by far the biggest of those at 12.3% (8,944 songs). Then come Cyrillic (4.3%), Indian scripts like Devanagari, Bengali and Tamil (2.1%), and Chinese, Japanese or Korean (1.6%).

Most English-language coverage of AI music focuses on English-language songs. Meanwhile, one of the biggest groups of creators in our data writes in Arabic. The Latin-alphabet share hides plenty of variety, too: Spanish, Portuguese, Indonesian, French and German prompts all use it.

The sounds people ask for

We read each song's style description and placed it in one genre family. The biggest group, at 16.9%, is music described as classical, orchestral or cinematic. That's piano pieces, film-style scores, choirs and big string arrangements.

After that it's a crowd:

• jazz and blues (8.1%)
• pop (6.7%)
• R&B and soul (6.4%)
• gospel and worship (5.9%)
• rock (5.7%)
• folk and acoustic (5.6%)
• hip-hop (5.3%)
• electronic (5.2%)
• Latin (5.1%)

Put hip-hop together with trap and drill, and rap reaches 8.6%.

Smaller scenes have real followings, too: afrobeats (2.5%), country (1.8%), K-pop (1.2%), phonk (1.0%) and lo-fi (0.9%). People use AI to make the music they already love, not just whatever is easiest to generate.

More people bring their own lyrics

You can start a song on MusicWave with a one-line idea or with a full set of lyrics. In July, 22.6% of songs started from lyrics. By October it was 37.3%, more than 1 in 3.

Some people write every word themselves. Others draft with a tool like our AI lyrics generator and then edit. Either way, more creators want a say in exactly what the song says.

Vocals tell a similar story. In March, 65.5% of songs had vocals; since May that share has stayed between 83% and 86%. Among the 1,598 songs where the creator picked the singer's voice, male voices were chosen about twice as often as female ones (1,080 to 514).

Creation keeps speeding up

In March 2026, people made 2,431 songs on MusicWave. In October they made 26,664, about 11 times as many. The number of people making songs each month grew from 712 to 6,470 over the same stretch.

This is one platform's slice of a much bigger category. Suno said in February 2026 that it had passed 2 million paid subscribers and $300 million in annual recurring revenue. To compare the tools themselves, start with our guide to the best AI music generators.

Small findings we didn't expect

• 1 in 4 songs runs past four minutes. The average is 3:22 and the median is 3:18, so these are full-length songs, not clips. 3.6% run past six minutes.
• 93.6% of songs are brand new. Only 6.4% start from existing audio:
• remixes (4.0%)
• added vocals (1.1%)
• extensions (0.8%)
• covers (0.3%)
• replaced sections (0.2%)
• Thursday is the busiest day. Across 34 full weeks, Thursday beat Tuesday, the quietest day, by about 16%.
• 1 in 5 songs gets downloaded. People keep what they make.
• 1 in 15 songs (6.8%) started from one of our example prompts. We left those out of the theme numbers so the themes reflect what people wrote themselves.

How we did this

• The data: all 72,637 finished songs still in creators' libraries on MusicWave.ai, made by 21,794 people between October 12, 2025 and October 30, 2026. Songs that creators deleted aren't included.
• Themes: we matched keywords in seven languages (English, Spanish, Portuguese, French, German, Russian and Arabic) against each song's prompt and title.
• Shares use the 45,813 songs whose creators described the song in their own words.
• Left out: 21,856 lyrics-first songs, which have no description, and 4,968 songs made from example prompts (any prompt used word for word by 20 or more people).
• Genres: read from each song's style description, one family per song. 12.8% of songs had no clear genre.
• Writing systems: detected from the Unicode script of each prompt and title.
• Privacy: everything here is aggregate. We didn't read, quote or share any individual prompt, lyric or account.

We'll refresh these numbers as the library grows.

Press and reuse

Every chart and number in this study is free to republish, in whole or in part, with credit to MusicWave.ai and a link to this page. For high-resolution charts or a custom cut of the data (by month, script or genre), email hello@musicwave.ai.

Want to make one yourself?

The data says that, for a lot of people, an AI song is a way to say something to someone. If there's a person you'd write one for, MusicWave can turn your memories into a personalized song with their name in it, ready to share as a link or an MP3.

FAQ