AhlanAI

An Arabic-first directory for discovering AI tools — 500+ tools across 20+ categories, with the Arabic experience designed as the primary one rather than a translated afterthought.

Overview

AhlanAI is a directory for discovering AI tools, built for the Arab market. It covers more than 500 tools across more than 20 categories, and it is fully localised — not in the sense of having an Arabic toggle, but in the sense that the Arabic version is the one the product was designed around.

It is live at ahlanai.com.

The problem

Anyone searching for AI tools in Arabic runs into the same wall. The good directories are English-only. The Arabic results are machine-translated listings with descriptions that read like an API response, categories that don't match how anyone here would group things, and search that fails the moment you type a word with a different hamza.

The result is that Arabic-speaking users default to searching in English — which works if your English is strong and quietly excludes a lot of people if it isn't. That's the gap: not a missing translation, a missing product.

My role

Founder and builder. Product definition, information architecture, the Arabic-first design decisions, the build, and the content model behind the catalogue.

This is a product I own rather than a client engagement, which means the decisions on this page — including the ones that turned out to be wrong — are mine.

What was built

A browsable catalogue of 500+ AI tools organised into 20+ categories, structured around how people actually look for tools rather than how vendors describe themselves.

An Arabic-first interface: RTL as a structural layout property, Arabic typography set independently of the Latin scale, and copy authored in Arabic rather than translated into it.

Arabic search with normalisation for the orthographic variants that break naive implementations — hamza and alef forms, taa marbuta, diacritics and elongation.

A parallel English experience that is genuinely first-class rather than the fallback, since most users move between both languages.

A content model that keeps tool descriptions, categories and metadata maintainable in both languages without one drifting out of sync with the other.

The Arabic-first decisions

The interesting engineering here isn't the directory — it's what "Arabic-first" costs and buys. Building RTL structurally rather than as a set of overrides is slower for the first three screens and faster for every screen after. Choosing Arabic type properly means the Latin and Arabic scales are tuned separately, which doubles a design decision that most projects make once.

Search was the decision with the clearest payoff. Arabic has several ways to write the same word depending on hamza placement, whether diacritics are included, and whether the writer used taa marbuta or haa. A user typing any of those variants expects the same results. Without normalisation, a directory returns nothing for queries that are plainly correct — and users conclude the product is empty rather than that the search is broken.

The general lesson, which now shapes every bilingual build I take on: the Arabic decisions that matter are made in the data layer and the type stack, long before anyone writes Arabic copy.

Stack

Only what was actually used:

Built with Lovable as the development platform.

React and TypeScript on a modern front-end toolchain.

Supabase for data and content storage.

A bilingual content model covering catalogue entries, categories and metadata.

Outcome

The product is live and serving an Arabic-first catalogue of 500+ tools across 20+ categories. Traffic and engagement figures aren't published here — unverified numbers on a case study are worth less than the absence of them.

What it does demonstrate concretely: an Arabic-first AI product, defined, built and shipped by one person, that treats Arabic as the design target rather than a localisation task.

AhlanAI — Visit the live product

Arabic-first AI Products — The approach, in detail

AI Product Development — How a build runs

Al-Qabas case study — Platform work at scale

All work — Selected projects

Building for an Arabic-speaking audience?

The decisions that matter get made before the first screen is designed. Worth having that conversation early.