Recent updates

Analyzing September's genre shake-up

Two big things changed on SubmitHub in September:

Both of them shifted which genres are being submitted. For a deep dive, you can visit the Genre Analytics page.

Highlights

The short version: the number of human-made songs barely changed. What changed is where those songs landed, and how much AI came along with them.

Normalizing for the # of days in the month, non-AI submissions were up +2.8% over August, while AI-flagged songs were down ~54% (this is the "85%+ AI" cutoff at work). So, curators seeing a change in their submission volume last month were likely impacted by either the genre re-shuffle or AI being dropped. The solution (if you're so inclined) is to focus on genres you like that have shown positive trends under the new genre model.

Oh, and one important note: these genre numbers aren't final. We're still tweaking and improving the genre model as we get more user feedback. So the model for October might produce some slightly different numbers!

The 20 biggest growers

  • Dark Pop: +72% (342 to 588 songs)
  • Alternative Dance / Stutter House: +67% (280 to 467)
  • Hyperpop: +44% (233 to 336)
  • Commercial: +33% (630 to 840)
  • Lo-fi Rock: +29% (259 to 334)
  • Red Dirt: +28% (247 to 317)
  • Electro Pop: +26% (601 to 755)
  • IDM / Glitch: +21% (228 to 276)
  • Grunge: +13% (214 to 242)
  • Afrobeats / Afro-pop / Afro-fusion: +11% (613 to 682)
  • Indietronica: +11% (549 to 607)
  • Americana: +10% (747 to 825)
  • New age: +10% (286 to 314)
  • Jazz Fusion: +10% (205 to 225)
  • Indie Dance: +9% (394 to 431)
  • Emo: +9% (398 to 435)
  • Post Rock: +8% (279 to 300)
  • Funk: +7% (262 to 280)
  • Cloud Rap / Emo Hip-Hop: +6% (473 to 503)

Most of these had been flat for a year before jumping in September, which tells us this is the new detector at work and not a sudden wave of Dark Pop, for example. It's better at picking the specific niche a song belongs to instead of dropping it into a broad bucket, so the more precise genres are filling up.

Smaller genres that more than doubled

These shifts look big, but are also the most-likely to change as we continue to tweak our genre model.

  • Modern Progressive Rock: +184% (127 to 361 songs)
  • PluggnB / Plugg: +164% (53 to 140)
  • Chillstep: +153% (53 to 134)
  • Slacker Rock: +137% (108 to 256)
  • Jangle Pop: +137% (63 to 149)
  • Psychedelic / Freak Folk: +125% (68 to 153)
  • International Hip-Hop: +117% (59 to 128)
  • Neofolk: +116% (55 to 119)
  • Folktronica: +115% (52 to 112)
  • Electronicore: +108% (66 to 137)
  • Folk Punk: +104% (50 to 102)
  • Lo-Fi Rap: +102% (111 to 224)

The 20 biggest drops

Many of these are still very active genres. Just less so than before. So don't rush to disable them outright!

  • Electro: -52% (301 to 143 songs)
  • House (Old-school): -49% (1,477 to 757)
  • Dance Pop: -43% (2,050 to 1,165)
  • Synthpop: -42% (1,041 to 601)
  • Country: -39% (772 to 473)
  • Pop Rap: -37% (1,092 to 687)
  • Reggae: -37% (259 to 163)
  • Gangsta Rap: -37% (283 to 179)
  • Contemporary R&B: -35% (1,142 to 739)
  • Reggaeton: -33% (374 to 252)
  • Trap (Hip-hop): -32% (1,376 to 932)
  • Pop Rock: -32% (2,008 to 1,365)
  • Techno: -32% (581 to 397)
  • Latin Pop: -28% (484 to 349)
  • Bass House / Electro House: -28% (773 to 558)
  • Deep House: -27% (1,474 to 1,070)
  • Dubstep: -25% (211 to 158)
  • Cinematic / Epic Music: -25% (701 to 525)
  • Hard Rock: -24% (582 to 444)
  • Punk: -23% (434 to 336)

Two things are going on here. The first is the flip side of the list above: broad catch-all genres like House (Old-school), Synthpop and Electro lost songs to more specific genres. As I pointed out when rolling out the new model, many of these genres were appearing when they shouldn't (eg, House (Old-school) coming up for basically every single modern house track).

The second is the AI ban. Dance Pop lost 885 songs, and 534 of them were AI. Pop Rock lost 643, of which 376 were AI. For Country it was 183 out of 299, and for Contemporary R&B 228 out of 403.

What happened to the AI

Across all genres, songs flagged by our AI detector fell 57% from August to September. In August, about 14% of submitted songs were flagged as likely AI. In September it was 6.5%.

Some notable changes:

  • Adult Contemporary: 36% to 14%
  • Christian: 28% to 10%
  • Country: 35% to 18%
  • Contemporary R&B: 26% to 10%
  • Bass House / Electro House: 32% to 16%
  • Retro Soul: 23% to 7%
  • Latin Pop: 30% to 14%
  • Pop Rock: 26% to 11%
  • Singer Songwriter: 22% to 7%
  • Dance Pop: 41% to 27%

The genres that still see the most AI:

  • Dance Pop: 27%
  • Country: 18%
  • Trance: 17%
  • Commercial: 16%
  • Bass House / Electro House: 16%

And the ones that see almost none:

  • Shoegaze: 0.4%
  • Psychedelic Rock: 0.6%
  • Experimental Electronic: 1.0%
  • Ambient: 1.5%
  • Punk: 1.5%

So if you make shoegaze, your competition is almost entirely other humans. If you make dance pop, it's a lot more human than it was a month ago.

You can dig into any genre, including its month-by-month history, on the Genre Analytics page.