Joe Doveton
About
Joe Doveton is the founder of GEO Jetpack, a generative engine optimisation (GEO) auditing platform that works with enterprise brands, including Lloyd’s Register, The Open University, Weatherbys Private Bank, and NOW: Pensions. He is also a director at Binary Bear, a business-to-business (B2B) consultancy specialising in search engine optimisation (SEO) and GEO.
Joe has spent over two decades in digital marketing, having previously built and sold the software-as-a-service (SaaS) product GlobalMaxer. He is a Chartered Institute of Marketing (CIM) trainer and a regular speaker at European search marketing events, including BrightonSEO. His work focuses on the intersection of entity-based SEO, artificial intelligence (AI) search visibility, and practical changes to content architecture that determine whether brands appear in answers generated by ChatGPT, Claude, Gemini, and Perplexity.
Joe Doveton | Building a GEO Auditing Platform with AI: 12 Months of Lessons from a Solo Founder
Twelve months ago, Joe Doveton - a creative director and search engine optimisation (SEO) consultant with no formal engineering background set out to build a generative engine optimisation (GEO) auditing platform from scratch. Today, GEO Jetpack is in production with a dozen enterprise clients, including Lloyd’s Register, The Open University, and Weatherbys Private Bank. Joe built the entire platform without writing code directly, using artificial intelligence (AI) as his primary development partner for architecture, debugging, deployment, and feature design.
This session provides an honest account of what that journey was actually like. Joe walks participants through the technical architecture. Node.js and Express on Railway, React on Vercel, PostgreSQL, Clerk for authentication, and integrations with four large language model (LLM) APIs, alongside Google Knowledge Graph, Wikipedia, and Wikidata. He covers production incidents that nearly broke the platform, including a single-character typo in a brand name that resulted in 0% visibility scores for an enterprise client, a PostgreSQL disk-full incident that manifested as silent audit failures, and a shape-mismatch bug that took four hours to resolve and involved five distinct root causes.
The session also addresses the strategic decisions that shaped the platform: pricing iterations, the sandbox model for client onboarding, why a free entry-level product was repositioned as a paid product, and how AI-assisted development changes who can build production software.
Attendees will leave with a realistic picture of what solo, AI-assisted software-as-a-service (SaaS) development looks like in 2026 - the genuine capabilities, remaining limitations, and workflows that turn AI from a code-completion tool into a development partner.