UAEBanks.org - UAE banking comparison built from messy public data

1) Why we built it

UAE banking information is public, but not easy to compare. A single bank may publish cards, accounts, finance products, fees, Key Fact Statements, terms, and promotional pages across many URLs. Product names are often repeated differently across pages and PDFs, and useful values such as salary, annual fee, foreign transaction fee, lounge access, cashback, miles, and required documents are rarely presented in a common structure.

• User need: people want simple answers before applying for a card, account, or loan.
• Business need: source-backed data should be reusable for comparison, eligibility, and lead workflows.
• Engineering need: the website had to stay fast, static, affordable, and easy to deploy.

2) Data collection and normalization

We started with a bank directory, verified bank websites, and then built crawlers and custom scrapers for bank product sections. Some banks exposed structured JSON, some required rendered HTML, and some only revealed product details inside individual pages or linked documents.

3) PDFs, KFS files, fees, and terms

Banking products often hide the most important information inside Key Fact Statements, schedules of charges, tariff PDFs, terms and conditions, and downloadable brochures. We treated documents as first-class sources rather than secondary attachments.

• Lightweight triage identifies PDFs likely to contain product, fee, eligibility, or terms data.
• Document extraction converts useful text and tables into Markdown/JSON-style research outputs.
• Relevant source links remain attached to products so users can verify bank-published documents.
• Non-useful crawl artifacts are excluded from the deployable site to keep the static build practical.

4) AI enrichment and review controls

LLM extraction was used where deterministic parsing could not reliably convert page sections into reusable facts. The goal was not to invent data; it was to turn visible source text into structured values, bullets, eligibility rules, fees, documents, benefits, and evidence snippets.

• Key-value facts: salary, annual fee, profit rate, repayment period, documents, and limits.
• Comparison fields: normalized fields that can drive filters and sorting.
• Evidence: source URLs and document links remain attached for review.
• Cost control: enrichment jobs use batches and hard cost caps instead of uncontrolled crawling.

5) Product UX, comparison, and eligibility

The UI is built for users who may not know banking terms. Product cards show short names, bank logos, product images, essential facts, benefit icons, and clear actions. Product pages explain benefits, fees, eligibility, documents, source links, and contact/report actions without exposing internal data scores.

• Credit card pages support comparison, benefit icons, card images, and official source links.
• The eligibility modal asks simple questions and can later connect to a lead workflow.
• Report-info forms allow users to flag missing or incorrect information.
• WhatsApp contact prompts connect high-intent users to human help.

6) Static pages, SEO, AEO, and AI discovery

UAEBanks.org uses a static build so every bank, service, and product can have a clean URL, metadata, sitemap entry, FAQ structured data, and AI-readable context. This keeps hosting simple while still making hundreds of product pages discoverable.

• Cloudflare Pages can deploy the generated static output.
• Product URLs use readable routes such as /products/credit-cards/product-id/.
• The build generates sitemap, robots, llms.txt, llms-full.txt, and MCP discovery files.
• Curated images are committed separately from bulky crawl caches so deployment stays reliable.

7) Technical breakthroughs

The hardest part was not building a static website; it was making a repeatable data system from inconsistent bank websites. The project combined bank-specific scrapers, general crawlers, PDF research, image curation, LLM-assisted extraction, sanitizer rules, static rendering, and user-facing comparison UX into one deployable product.