Photozilla: paid ads experiment #1

Hi, I hope this email finds you well. I'm trying a new kind of post for my blog: learnings from the actual things I do. When I do concrete stuff about my projects, put them out there and get some data I can share, I am thinking to do it in this format. Let me know how you feel about this approach by replying at this email (if you got it through in your inbox) or at vlad@vladcalin.ro otherwise.

I ran a little paid traffic experiment (on Google Ads) for photozilla.io in the past few days. It was... weird.

Most of the traffic that came in was from the "ai images" keyword family, and not from the "stock photo" one, as I was hoping. This is an indicator that there is a strong demand for AI image generation, but people looking for stock photos are avoiding AI products in that niche. And that strong AI image generation demand probably means that it can be monetized to some degree, but here's where it gets interesting:

• almost all new accounts (~30 of them) came from @gmail.com accounts with human names, meaning they are consumers, and not businesses)
• there were a lot of generation attempts for porn and naked people. And plenty some weird ones with anime characters in diapers. These were some data points I had not anticipated.
• there were some attempts to generate straight illegal content, meaning that if you build something in the B2C niche with a user generated content component, you have to invest in content flagging and moderation from day one.

With these in mind, I'll have to see where I can go with Photozilla. I have a few leads:

• there was another idea I had before it, an API/service to extract structured data from images. I could move in that direction.
• add a layer on top of it to aid in generating social media ready visuals (add support for logo, text placement).
• pursue one of the leads in the prompts and niche down (one is "christmas backgrounds").