Overview & result
JevSEO analyzes a URL or draft against separate SEO, answer-engine and generative-engine rubrics. Code measures fixed signals and composes the scores; Jev handles text-dependent judgments. A site-wide mode crawls pages, proposes actions and flags overlapping intent.
From the original post
Score a page for search, answer engines and generative engines — with judgments you can audit.
How Jev fits in the loop
- Fetch and parse a page, then compute deterministic signals such as measurable counts and blocking conditions in code.
- Ask Jev the applicable semantic rubric questions in a batch and retain the answers and confidence for review.
- Compose separate SEO, AEO and GEO scores and rank fixes from the recorded rules and judgments.
- In site-wide mode, crawl same-origin pages and propose typed leave, improve, rewrite, merge, split or prune actions.
How to reproduce
- Clone the public repository, run pnpm install and pnpm dev, and add a TypeSafe API key through its local settings page as the README directs.
- Use /analyse with a URL or draft to inspect the three pillar scores, ranked fixes and supporting judgments.
- For site-wide work, use /projects and inspect the author-described leave, improve, rewrite, merge, split and prune recommendations before acting.
Why this build matters
The public repository makes the boundary between rules, semantic judgment and scoring policy inspectable, including when a low-confidence answer needs human review.
Limitations
- SEO, AEO and GEO scores are the project’s own rubrics; they are not Google ranking or citation predictions.
- The speed and cost figures in the README are author-reported and were not benchmarked by JevForAgents.
- Live SERP comparison is optional and requires separate DataForSEO credentials; it is not part of the default local analysis.
For a different SEO use of Jev, compare the Ryze AI workflow demo. It describes product decisions in an original post; JevSEO has a public code repository.