Overview & Result
MyOrlandoStay.com reports using Jev with Devin to score 25 pages against the question ‘would an AI assistant find the answer here?’ Devin handled prerendering, JSON-LD, llms.txt, and IndexNow. The author reports SEO moving from 88 to 97.6 and AI-EO from 43 to 86.
How Jev fits in the loop
- Jev evaluates each page against a bounded AI-assistant answerability question.
- The engineering agent applies discoverability changes such as prerendering and structured metadata.
- The site owner compares the reported score before and after the changes.
How to reproduce
- Open the original post and record the author's definition of the assistant-visibility score.
- Create a fixed page set and a bounded question such as whether an AI assistant can find the answer.
- Keep Jev scoring separate from the engineering changes, and label any before/after result as author-reported until independently measured.
Why this build matters
Use a typed evaluator to prioritize which pages need better machine-readable answers.
Limitations
- The post does not disclose the full rubric, prompts, thresholds, or page URLs.
- The before/after scores are not a substitute for Search Console impressions, clicks, or independent AI-answer tests.