Rebuilding a slow site into a static-first platform AI search can quote
Turning a sluggish, framework-heavy marketing site into a static-first Astro build that hits 100/100 Core Web Vitals and is structured to be quoted by AI answer engines.
- 100/100
- Core Web Vitals
- Sub-1s
- largest paint on mobile
- Cited
- in AI answers
The problem
The site ran on a heavy client-rendered stack. It loaded slowly on real devices, failed Core Web Vitals in the field, and was effectively invisible to AI answer engines: nothing on the page was structured for a machine to read, trust, or quote. Traffic was flat and the brand never surfaced in AI-generated answers.
What we did
We rebuilt it static-first on Astro, shipping almost no JavaScript by default and tuning the critical path so the largest content paints quickly and the layout never shifts. Then we structured the content for Generative Engine Optimization: answer-first sections, named entities, JSON-LD describing the business, and a machine-readable summary, so AI engines can extract and cite it rather than skip past it.
Techniques
Astro static-first Google's three headline measures of how a page feels: how fast the main content appears, how quickly it responds, and whether the layout jumps. Interaction to Next Paint. How long a page takes to visibly respond after a tap or a click. Generative Engine Optimization. Structuring a page so ChatGPT, Perplexity, and Gemini can read it, trust it, and quote it in an answer. A structured description of what a page means, written for machines. It is how search and AI engines know what they are looking at.
Why the old build failed twice
The site failed two audiences at once. For people, a client-rendered framework delayed the largest paint and shifted the layout as scripts hydrated, so it felt slow on real phones. For machines, the meaningful content was assembled in the browser and carried no structured data, so search and AI engines had little to read and nothing authoritative to quote.
Fast by construction, not by patching
Performance became a property of the architecture rather than a later optimization pass. Static-first Astro ships finished HTML with almost no JavaScript, fonts are preloaded to remove layout shift, media is sized and served from an edge CDN, and the critical path is kept short. The result is 100/100 Core Web Vitals that holds on real devices, not just in the lab.
Built to be quoted, not just ranked
Being citable by AI is now a separate discipline from ranking: the overlap between top search links and AI-cited sources has fallen sharply. We wrote answer-first sections that stand alone, named the concrete entities the business should be associated with, emitted JSON-LD for the organization and its pages, and published a machine-readable summary, so an AI engine can lift a correct, attributable answer.
Questions on this project
Does Generative Engine Optimization hurt traditional SEO?
No. The same structure that makes a page extractable for AI, clear headings, answer-first sections, valid structured data, and fast load, also strengthens classic SEO. GEO adds machine-readable context on top; it does not trade away rankings.
Why static-first instead of a JavaScript framework?
A static-first build ships finished HTML, so content is fast for people and readable by machines without running any code. Heavy client rendering delays the largest paint and can hide content from engines that do not execute JavaScript.
How do you measure whether AI engines actually cite a site?
By tracking referral traffic from AI engines such as ChatGPT, Perplexity, and Gemini, and by checking whether the site appears as a named source in their answers for target questions, alongside the usual Core Web Vitals and search metrics.