I built a $100/month social media intern for my directory
written by Frey Chu
I built a $100/month social media intern for my directory
This week I did something I've been putting off for a while. I tried to figure out if there's a way to drive traffic to my directories beyond SEO.
I love SEO. It's the foundation of everything I've built. But I've been very reliant on it as my only traffic channel, and creating social media content every day has always felt like too much work for a directory site.
Then OpenClaw started blowing up, and I figured it was time to actually test it.
The Inspiration
My friend Meg Kuhn runs a community called Women Building with AI, and she created a TikTok account to promote it.
Every single post is just a simple photo carousel, nothing fancy, and she's consistently getting thousands of views. A percentage of those views turn into people clicking her bio and discovering her community.
I looked at that and thought, what if I could build something like this for my directories, but have AI do the entire thing? The content creation, the captions, the scheduling, all of it. No manual work from me.
Meet TikTok Tom
So I built an AI agent called TikTok Tom. Twice a day it wakes up, pulls real pricing data from my DiscoverPlasma Supabase database, generates TikTok photo carousels using templates I designed, writes a caption with real data and hashtags, and schedules the post through a tool called Postiz. It even tracks which listings it has already posted about so it never repeats content.
The whole thing runs on its own. I just check in once a day to see what it posted (though right now, I'm trying to make the content better before actually running it 24/7. More details below).
Here's how it works at a high level. There are four pieces to the pipeline:
The database. I already had all my DiscoverPlasma data in Supabase. 1,250 plasma centers with crowdsourced pricing data. OpenClaw gets read only access to pull from it.
The templates. I designed TikTok slide templates in Claude Chat that look like Apple Notes screenshots in dark mode. The layout stays the same, but the center name, city, and pricing numbers swap out automatically per listing.
One of four photo carousels that make up the tiktok "content" my agent created for me
The brain. That's OpenClaw. Every day it picks a listing, plugs the data into the template, renders the slides, writes the caption, and sends it to Postiz.
The posting. Postiz handles the TikTok connection so I didn't have to deal with TikTok's developer API, which requires an application that could take weeks and you might even get rejected.
Is OpenClaw friendly to non-coders?
I'm not going to sugarcoat this part. I spent about three hours just getting OpenClaw installed, and then another four to six hours trying to get it to do useful work.
I started with ChatGPT and it was pretty bad. It kept describing plans instead of executing them and would go down rabbit holes assuming things I didn't want. When I switched to Claude Opus 4.6, everything clicked. It understood the full picture, executed immediately, and I was able to get the whole pipeline working within a day.
The model you connect to OpenClaw matters way more than OpenClaw itself. That was probably my biggest takeaway from the whole experience.
- Paid Partnership -
If you don't want to be like me and spend three hours configuring OpenClaw from scratch and a hundreds dollars on a Mac Mini, Hostinger gives you a virtual computer in the cloud for about seven bucks a month.
They built a one click setup specifically for OpenClaw, so instead of spending hours figuring out how to install everything yourself, you click one button and OpenClaw is ready to go. Your environment is fully isolated and secure, and your agent stays online around the clock even when your computer is off.
Photo carousels aren't exactly the most engaging format on TikTok. But the play here isn't viral content. It's TikTok SEO.
Every post targets a specific city and plasma center. So when someone in Houston searches "plasma donation Houston" on TikTok, my post is sitting there. Over time, hundreds of location specific posts create a long tail search net across every city I cover.
At least, that's what I'm hoping will happen.
I'm also exploring ways to improve the content quality by layering on tools like ElevenLabs for AI voiceovers and Shotstack for automated video assembly. That's the next evolution of this experiment.
Should you try this for your directory?
I think it depends on your niche and how you monetize.
For most display ad directories, you'd need a ton of social traffic to justify the $100/month it costs to run.
The math is tough if you're not getting over 10k/monthly visitors consistently.
For a lead generation directory where one lead is worth hundreds of dollars, the math has a chance to work out much better. My luxury restroom trailer directory would be a way better candidate for this than DiscoverPlasma right now.
And if you're building in a niche that's already popular on TikTok, like thrifting, food, fitness, or real estate, the audience is already there. You're just showing up with answers to questions they're already searching for.
Side-note: This could be a service you sell. A fully autonomous social media agent that posts every day for a flat monthly fee. Your cost is about $100/month. What you charge is up to you.
I'm giving you the full blueprint
If you want to create your own autonomous social media intern, I'm giving you what I spent 40 hours creating.
After I got everything working, I asked OpenClaw to create a folder with a setup guide, skill.md and other markdown files that walks through the entire setup. Accounts, database mapping, template design, daily automation, troubleshooting, all of it.
If you want to build your own version of TikTok Tom, this should save you a lot of the time and trouble I went through.
Start with tiktok-tom-getting-started.md. This is your overview of the entire pipeline, what tools you need, and how to adapt it to your niche.
Then read OPENCLAW-INSTRUCTIONS.md. This is the step by step setup walkthrough. Follow this when you're ready to build.
Next is SKILL.md. This is the full technical reference and the most important file. This is what your AI agent reads to run the pipeline. Once you've adapted it to your niche, feed this to OpenClaw and it knows everything.
FALLBACK-LOGIC.md covers how the system picks templates based on what data a listing has. Read this when you're designing your template tiers.
openclaw-supabase-tables.md and openclaw-center-pay-summary.md are database schema references. Use these as examples when mapping your own Supabase tables to the pipeline.
Where I go from here
SEO is still my primary strategy for directories and I don't see that changing. This is more of an experiment.
But a $100/month experiment that builds a second traffic channel on autopilot while I focus on everything else? I think that's worth running for a while and seeing where it goes.
