The economics of an AI UGC phone farm

So you're thinking of marketing your app on social media with organic UGC. You have a few options to pick from:

• influencer marketing (reaching out to people who have followers to mention your app)
• "canvas" UGC (having people create fresh new accounts on social media to post videos about your app)
• AI UGC (creating AI videos that mention your app, and posting them usually yourself)
• and more

In this article, we will make the case for posting AI UGC yourself, ie. setting up (and ideally automating) your own phone farm to post AI UGC. We will compare that against the currently most (?) popular option, human canvas UGC.

The economics

The unit for comparison we will use is CPM, ie. cost per thousand views, on your videos.

The typical cost for human canvas UGC involves a $500 monthly subscription to UGC creator marketplaces (such as SideShift), which we will charitably ignore though this can be a heavy burden on new projects. The benchmark rate excluding this fixed cost you will often see quoted is ~$1 CPM. Different payout schemes can bring this number down (such as setting a maximum payout per video), but it is rare to see figures that are significantly lower than $1 CPM, at scale.

The cost for AI UGC breaks down as per below - in this breakdown, we will not exclude any cost:

• cost of a used iPhone: $100
• cost of a residential dedicated proxy: $4 / month
• cost of an SMS verification service: $0.05 (textverified dot com)
• cost of an AI image (for slideshows): ~$0.05, cost of an AI video: ~$0.8. (for 10-15 seconds, good workflows can bring this down). We'll assume you do a 50/50 mix of 6-slide slideshows, and videos, for an average of $0.55 per piece of content

Nothing else is needed to get you started. Considering the following:

• a used iPhone can be used to post 3x / day, on 5 accounts, safely (ie. 15 posts / day / iPhone)
• (optional) on top of this, the same content can be posted across multiple platforms (TikTok, Instagram, YouTube Shorts, Snapchat, Facebook, Threads)
• assuming an average of 1,000 views per slideshow or per video (easily achievable if the content isn't terrible)

Putting all of this together,

• Costs on month 1 = used iPhone (100) + proxy (4) + SMS (0.05) + slideshows / videos (247.5, based on 15 posts / day for 30 months at $0.55 each) = $352
• Views on month 1 = average views per content (1,000) * 450 (assuming you ONLY post on TikTok) = 450,000
• CPM = $0.78

The most charitable projection of costs for canvas UGC, against the least charitable projection of AI UGC, makes AI UGC about 20% cheaper. But this, of course is not the real strength of AI UGC. The real savings kick in on month 2, then month 3, etc., as your fixed cost (iPhone) drop in its contribution to its total cost. On top of that, posting on more social media platforms (up to 6 platforms) further divides the costs to a fraction. All in all, posting on 3 social media websites, by month 12, your CPM is closer to $0.08, ie. more than 90% cheaper than dealing with human UGC creators.

In my own (optimized workflow, mainly slideshows), my CPM is closer to $0.03. For other options to make sense at $1 CPM, the delta in conversion rate literally has to be 30x better.

The management, and other considerations

The delta in costs is clear, but posting several times a day, swiping the FYP for a few minutes before & after posting on a daily basis, morning, noon, and night, is a real burden. The obvious solution then becomes automating a phone farm. Having tried both options above, my sense is that:

• Starting human UGC is less work / management overhead in the beginning (setting up a good, undetectable phone farm takes a bit of work)
• After the initial setup, a phone farm requires significantly less maintenance

My personality tends to favor real autonomy, and reject or shy away from high-touch human management overhead. I'd rather improve my automatic setup.

Another consideration you may want to incorporate is that of future-proofing your setup. AI UGC is now clearly at the stage where well-executed campaigns are extremely successful (and high conversion). We have many such examples. Now imagine the cost, speed, and quality we'll have next year (insert will_smith_eating_spaghetti.mp4).

This is the part where I shill my approach. As a result of the above, I invested all of my efforts in finding a path to automation that is undetectable. No APIs (RIP reach), WDA (good luck making WebDriverAgent undetectable and injecting a radius to your gestures), no OTG (pesky isCaptured). I wanted a long-term sustainable, low maintenance, cheap and scalable source of massive views. The results of several months of this effort is ghostfarm.io. At $99 / month, the impact it has on CPM is +$0.07.

If this vision on how to operate an online business compels you, join the ghostfarm.io discord and learn from the community. Follow me on X for more tips & tutorials on setting & scaling this up.

We don't sow anymore. We just reap.