We analyzed 340 faceless voiceover YouTube channels across 25 niches, most of them 100% AI-generated, to figure out what actually works and what's a complete waste of time.
Headline findings
Six questions. Data-backed answers. No vibes.
1) Pick your niche
Everything flows from this. Get it wrong and your thumbnail game, your AI pipeline, your posting schedule all become irrelevant.
"Start here" means high opportunity with low competition. Pay attention to the AI Mix column: 100% Full AI means everyone's output looks identical, while some Partial or Minimal mix signals room for someone with actual taste.
Scatter plot version where up-and-left = good and down-and-right = you're competing with 55 other channels for table scraps:
Everyone chases the biggest niche. Animation & Storytelling has 55 channels in our dataset and you're not going to out-grind 55 incumbents in month one.
Two niches alone eat nearly 30% of all channels:
Same niches, five different cuts. The last chart (payoff-to-competition ratio) matters most because it answers: "for each unit of crowding, how much subscriber upside do I get?"
2) How much AI should you use?
86% of channels in this segment are Full AI. That's descriptive, not prescriptive. The ones doing well got there early with first-mover advantage before the segment flooded with identical content, and shipping the same slop as 293 other Full AI channels in 2026 is not a strategy.
159 of 340 channels are stuck under 10k subs and 96% of them are Full AI. Meanwhile the 1M+ bracket is only 35% Full AI, with the rest using Partial or Minimal. The channels that got big have more human involvement. The ones stuck in the long tail automated everything and hoped the algorithm would carry them.
3) Where the AI headroom is
Forget which niches are biggest. The real question is which niches have proven subscriber demand but haven't been fully colonized by Full AI yet. If viewers are already watching and human-directed channels still dominate, that's your opening.
TheInfographicsShow: 15.3M subs, Minimal AI. coldfusion: 5.1M, Partial AI. melodysheep: 3.2M, Partial AI. These channels didn't automate everything and get lucky, they used AI as a tool and provided the taste themselves.
4) Your first 30 days
Stop reading articles, start publishing.
- Pick one niche from the leaderboard. Under 15 channels, over 100k avg subs. The data is right there.
- Publish 20 videos in one format. No pivoting, no "maybe I should try gaming." Twenty, same format. Most people quit after 5 and then post on Reddit asking why the algorithm hates them.
- Use Partial AI. AI generates, you curate. Write your own scripts or heavily edit the AI output, review every image, delete anything that smells generic.
- Track retention by topic cluster, not individual videos. One viral video teaches you nothing. Ten videos in the same cluster trending 2x above baseline teaches you everything.
- Automate more only after quality is predictable. Can't hold 40% retention to the midpoint? Then automating your pipeline just means you produce unwatchable content faster.
Full automation on day one isn't a channel strategy, it's a slop factory. YouTube's gotten very good at identifying slop and burying it.
5) None of this matters if you don't ship
Upload frequency is the strongest predictor of growth in our dataset. Not niche, not AI mix. The channels that broke 100k posted constantly for months without stopping.
The bottleneck is never ideas, it's production. Script, images, animation, voiceover, music, subtitles, assembly... each step takes time and each handoff is where your workflow dies. Most creators spend 4-8 hours per video on manual work that could be automated. At that rate you burn out in two months posting 2x/week.
OpenSlop fixes this. One prompt, one finished 15-minute video tailored to your niche, in minutes. Script, images, selective animation, voiceover, music, assembly in a single pipeline. You review, adjust, publish. That's the difference between 2x/week and daily.
The niche analysis tells you where to aim. The pipeline is what gets you there.
OpenSlop is the open-source workflow that creates ready-to-publish AI videos for free forever.
Join creators on the waitlist
Appendix: Deep dive charts
Not required reading. Useful if you want to argue with the data.
The power curve
Classic power law. A handful of channels have millions of subscribers, the vast majority have almost nothing. Niche selection gets you in the door, format and execution determine which side of this curve you land on.
Subscriber distribution by niche
The box is the middle 50%. If the average is 10x the median, that niche is winner-take-most: a few massive channels and everyone else fighting over scraps. You probably won't be one of the winners.
Total subscriber mass by niche
Art & Music dominates because of a few massive outliers. The interesting niches are the ones with big totals and few channels: demand without supply.
Supply vs. demand in one view
Marimekko chart. Column width = number of channels, column height = total subscriber mass. Tall and narrow = proven demand, almost no supply. Wide and short = crowded, low payoff. You want tall and narrow.
Full channel dataset
All 340 channels across 25 niches. Filter by category or AI usage level.
OpenSlop is the open-source workflow that creates ready-to-publish AI videos for free forever.
Join creators on the waitlist