Context
Today content has to show up in Google results and also in answers from AI assistants like ChatGPT. In the product, this second direction was called GEO. Usually both jobs are done by hand.
AI SaaS · 2026
VibeSEO
SEO content is usually made by hand, in a loop: collect keywords, write an article, publish it, start again. VibeSEO did this by itself. I co-founded the project, and as CTO I was responsible for the product and development.
What I built
A user signed in with Google, connected a site and got a keyword map. From that map the platform wrote SEO/GEO articles and published them to the site’s CMS by itself. I wrote plugins for Framer and WordPress. New users came through a free GEO audit: with no sign-up, it checked crawler access, schema markup, technical SEO and llms.txt.
Architecture
Back end on NestJS and TypeScript, data through Prisma, queues on BullMQ and Redis, front end on Nuxt 4. Keywords and competitors came from DataForSEO through MCP. Articles were written by a layer of several LLMs: OpenAI, Claude, Gemini and Perplexity. Publishing depended on someone else’s system, so an article counted as done only when the client’s CMS accepted it.
Challenges
Articles were created and published with no person involved, so I had to build checks and retries into every step. Also, no single model provider could become a hard dependency, which is why there was a layer of several LLMs.
Outcome
The platform ran in production until July 2026, then the project closed.
Results
Built the product and the teamIn production until 07.2026Project closed · 07.2026
Need to launch a product?
Send me the idea, the repo or the unfinished project. First I’ll ask questions, then I’ll send a written scope. Looking for someone to join your team? Tell me which role.