Website · 2026

Ant Arge Corporate Website

The bilingual corporate site of Ant Arge, a software development and R&D company, rebuilt for Core Web Vitals, technical SEO and AI search.

Year

2026

Type

Website

Role

Web Developer

Tech Stack

Astro
JavaScript
GSAP
Sharp
Satori
Docker
Cloudflare

Liveantargetech.com ↗

PrivatePrivate repository. Source code is not public.

Project Summary

The corporate website of Ant Arge, a software development, R&D and digital transformation company. It presents seven service areas, four case studies and a blog in Turkish and English, with Turkish at the root and English under /en. Every page is generated as static HTML at build time, so no framework JavaScript reaches the browser. The design started from a licensed ThemeForest template, adapted to the brand with the parts that slowed it down removed.

Most of the work went into speed and discoverability. Unused template CSS is purged at build time and inlined into each page, cutting it from about 720 kB to about 130 kB (around 30 kB gzipped). The mobile hero texture went from 700 kB to 40 kB, and images are converted to responsive WebP. Each page carries one schema.org graph, hreflang pairs and a canonical URL. The same data also generates llms.txt files for AI search engines. In mobile Lighthouse the home page scores 97 for performance and 100 for accessibility, best practices and SEO.

Problem & Solution

Problem

A render-blocking CSS file accounted for about half of the mobile LCP
The hero background texture was the LCP element, at 8.8 s on slow 4G
The template's smooth scrolling made the page feel heavy
Old blog and service URLs needed to keep their search value after the redesign
The site had to be readable by AI search engines as well as Google

Solution

Template CSS is purged at build time and inlined into each page
Responsive 640 px and 1200 px WebP textures with a preloaded hero image
Smooth scrolling was removed, and jQuery and the carousel load only on the home page
301 redirects for legacy English blog paths and old service URLs
llms.txt and llms-full.txt are generated from site data, and robots.txt allows AI crawlers

Key Features

Bilingual Routing

Shared route maps produce each Turkish and English page pair, with hreflang tags (tr, en, x-default) and both languages in the sitemap.

Structured Data

One schema.org graph per page covering Organization, WebSite, Service, SoftwareApplication, BlogPosting, FAQPage and breadcrumbs.

AI Search Files

llms.txt and llms-full.txt are built from the same data as the pages, so AI assistants get an accurate summary of the company.

Image Pipeline

astro:assets converts images to WebP. Build scripts render branded 1200×630 blog covers with Satori, plus responsive diagram images.

Blog

15 posts in each language, with internal links between related topics.

Privacy-Aware Analytics

GA4 loads only when the browser is idle or on first interaction, and stays off after a cookie rejection. The contact form submits with fetch, without a page reload.

Technical Details

Static Build

Astro 6 outputs about 80 static HTML pages without trailing slashes. The Docker image builds on Node 22 and serves the result behind Cloudflare.

CSS Budget

A PurgeCSS step strips unused template styles (about 720 kB down to 130 kB) and inlines the result, removing the separate stylesheet request.

Search Indexing

An IndexNow script pings Bing and Yandex with the live sitemap URLs after each release.

Audits

Fixes to contrast, heading order and link names for accessibility. A Cloudflare AI-bot setting that was lowering the SEO score was traced and corrected.

Previous ProjectAntApp WhatsApp Business Inbox Next ProjectDynamic Translator Product Website