Challenge
We regularly prepare content for the agency's SEO and GEO clients: articles targeting search queries, local pages and materials for AI answers. The process had several bottlenecks.
First, choosing topics. The best ideas are already in Search Console: queries the site appears for but gets few clicks on. Exporting them manually and grouping them by potential takes a long time.
Second, publishing. Clients use different CMSs: WordPress, Shopify, Strapi and custom systems. A separate publishing method for each means separate work and separate errors.
Third, choosing AI models. One model works better for one type of text, another for a different type, and images have their own services. Being locked into a single provider limits quality.
Finally, control. After publishing, you need to see whether the page got indexed and how it performs in Search Console and GA4. Without that, a content plan turns into a conveyor belt with no feedback.
What we did
We built the platform on Next.js 16, React 19, Tailwind CSS v4 and Supabase, deployed on Vercel. The first version took 3 days: 64 commits, 17 screens, 28 APIs and 20 tables.
Modules
- Topics: Search Console queries clustered by potential.
- Generation of text and images.
- Versioned editor for comparing and restoring previous drafts.
- Content plan with an automatic run every 15 minutes: scheduled pieces publish themselves.
- Analytics: GSC and GA4 data, indexing status.
- Integrations: GA4, GSC, the client's website, Google Business Profile.
- Team and AI keys.
One connector for different CMSs
Instead of separate logic for each website, we built a single connector that works with WordPress, Shopify, Strapi and webhooks. For a new client, it is enough to connect their website rather than write a new integration.
Model choice for every generation
The text provider (Claude, ChatGPT or Gemini) and image provider (OpenAI, Imagen or Replicate) are chosen separately for each generation. This way we match the model to the task, not to the platform's limitations.
Results
In 3 days, the agency got a working platform that covers the entire cycle: from a Search Console topic to a published page and monitoring of its indexing.
What changed in our work:
- topics come from real Search Console data and are grouped by potential;
- the content plan runs automatically, with scheduled pieces checked every 15 minutes;
- publishing to WordPress, Shopify, Strapi or via webhook goes through a single connector;
- indexing status and GSC and GA4 data are visible next to each piece;
- the model for text and images is chosen for the specific task.
We do not cite traffic numbers here, because this is a tool, and results depend on the specific website and strategy. SEO and GEO case studies where we work with content are described separately.
What makes this project special
Most content generation services are tied to a single model and a single CMS. Here, both limitations are removed: there is one connector for different platforms, and the text and image provider can be changed for each generation.
The second feature is feedback. The platform does not just publish; it shows whether the page got indexed and how it performs. This is what sets a content system for SEO and GEO apart from a text generator.
Finally, speed: 3 days from idea to a working version with 17 screens shows how we approach development for clients too.
Published: · Updated: