How to Test Arabic Support in an ATS Before You Buy

Every recruitment platform sold in the Gulf says it supports Arabic. Almost none of them mean the same thing by it, and the difference only surfaces after you have migrated your candidate database.

This is a test script. Run it during a demo, on your own data, and make the vendor drive. Each check takes a couple of minutes and each one has failed for a real product.

Four different things called “Arabic support”

Vendors answer “yes” to “do you support Arabic?” at level one. You need level three to run a pipeline and level four to trust your reporting.

The tests

1. Arabic CV parsing

Upload the four Arabic CVs. Then open each resulting candidate record and check the extracted fields, not the file preview. Many systems display the PDF correctly and extract nothing from it.

• Is the name populated, and is it the actual name rather than a header or a phone number?
• Are the job titles and employers extracted, or is the experience section empty?
• Are dates read correctly, including any written in Hijri?
• Do the skills appear as separate entries, or as one run-on string?

A specific failure to watch for: Arabic PDFs exported from some word processors store text in visual rather than logical order, so extraction produces reversed or disconnected letters. If the extracted name looks like scrambled characters, the parser cannot read Arabic PDFs and no amount of interface translation will fix it.

2. Search and matching

Search is where Arabic support usually fails quietly.

1. Search for a candidate by their Arabic name exactly as written on the CV. It should be found.
2. Search again without diacritics, and with a different but common spelling of the same name — محمد and محمّد, or a name beginning with a plain alif against one with a hamza. A system that treats these as different strings will lose candidates your recruiters know exist.
3. Search for an Arabic skill term and check that candidates who wrote it in English also appear, and vice versa. If not, every filtered shortlist you build will be half the real pool.

3. Scoring across languages

Post one job description in English, then screen an Arabic CV and an English CV describing equivalent experience. Compare the scores.

If the Arabic candidate scores materially lower for equivalent experience, the system is scoring language proficiency rather than fit — and you have just built a filter that disadvantages Arabic-speaking applicants at scale. Ask the vendor to explain the gap. Ask whether the score is computed on the original text or on a machine translation of it, because translation loses precisely the domain terms that scoring depends on.

4. Duplicate identity

Upload the same person twice: once from their Arabic CV, once from their English one. A system with bilingual identity handling will flag them as the same candidate or offer to merge. A system without it now holds two records, and every count you report afterwards is wrong.

5. What the candidate sees

Interface Arabic is for your team. Candidate-facing Arabic is what affects your application completion rate.

• Open the careers page in Arabic on a phone. Does the layout mirror, or only the text?
• Start an application. Are the form labels, validation errors and file-upload prompts all in Arabic, or does it fall back to English at the first error?
• Trigger a rejection or interview-invitation email. Which language arrives, and is the sender name readable?
• If there is an automated interview step, does it work in spoken Gulf dialect, or only in Modern Standard Arabic?

That last one matters more than vendors admit. Speech systems trained mostly on Modern Standard Arabic degrade sharply on how people actually speak, and a transcript full of errors produces an assessment full of errors.

Questions worth asking directly

• Does your CV parser process Arabic text natively, or translate it into English first?
• Does search normalise Arabic orthographic variants, and can you show me?
• Is your Arabic interface fully right-to-left, or translated strings in a left-to-right layout?
• How do you handle one candidate appearing in two scripts?
• Which parts of the product are not available in Arabic today?

The last question is the most informative. A vendor who can name the gaps has thought about the problem. A vendor who says everything works equally well in both languages has usually not tested it.

Reading your results

If a platform passes levels one and two but fails three, it is usable for an English-language pipeline with an Arabic veneer. That may be fine if your applicants apply in English. It is not fine if you are hiring at volume in the local market, because the failures compound silently: candidates who cannot be found, shortlists that are systematically incomplete, and headcount numbers that are quietly wrong.

Run the tests before you sign, not after. Our pricing is published, and the free tier exists so you can run this script on your own CVs without talking to anyone first.