Every paid media plan we inherit from a previous agency in the GCC has the same shape: 80% English-language creative, 20% Arabic creative, with the Arabic produced last, in fewer variants, and tested less. The reason given is always some combination of "Arabic is harder," "we don't have the writers," or "the dialect question is too complex."

All three reasons were genuine. None of them are still binding constraints in 2026. The Arabic creative bottleneck was a labour-supply problem, and labour-supply problems dissolve fast in an AI-native operating model.

Why Arabic was hard

Four structural challenges made Arabic creative slow and expensive in a traditional agency model:

Right-to-left layout. Arabic reads RTL, which means every layout, button position, image composition, and visual rhythm has to be rebuilt rather than translated. A Latin-script ad does not become an Arabic ad by swapping the words. Most agencies treated this as a final-step localisation task, which produced layouts that worked grammatically but felt off-balance to native readers.

Dialect variation across the Gulf. Modern Standard Arabic (MSA) is universally readable but emotionally flat. Khaleeji dialect plays well in the UAE, KSA, Kuwait, Bahrain, and Oman with subtle local variations. Egyptian dialect dominates entertainment but reads as out-of-place in a luxury hospitality ad in Riyadh. Levantine has its own tonality. Picking the wrong register is the kind of mistake that's invisible to a non-native marketer and obvious to the audience.

Cultural nuance. Arabic is denser with religious, social, and gendered cues than English. Word choices that are neutral in English carry implication in Arabic. Phrases that work in one Gulf market are awkward in another. Agencies dealt with this through senior-writer review, which created a serial bottleneck.

Type and typography. Arabic typography is genuinely different from Latin typography — kerning, ligatures, baseline behaviour, font availability. A Latin font does not have an Arabic counterpart by default; mixing scripts in a single layout requires deliberate font pairing that most design teams don't have a strong opinion on.

Stack those four constraints on top of the standard creative production pipeline and Arabic became the slowest, most expensive, lowest-volume part of every campaign. The result was a chronic under-investment that left brands with weaker performance in their largest market.

What changed

Three things shifted between 2024 and 2026 that, together, dissolved the bottleneck:

Models got materially better at Arabic. The base capability of large language models on Arabic — generation, dialect handling, register sensitivity — improved roughly 4–6× depending on the task. The 2026 generation handles Khaleeji, MSA, and Levantine with the kind of register awareness that previously required a senior native writer. They aren't perfect, but the failure modes are now reviewable rather than starting-from-scratch.

Specialised classifiers became deployable. We can now run a sentence-level dialect classifier and a register classifier on every piece of Arabic creative output. The classifiers were research artifacts in 2023; they're production tools now. Combined, they catch nearly all of the previously-invisible "this would never be said in this dialect" errors before the variant reaches a human.

RTL layout primitives matured. The visual generation models now handle RTL composition natively rather than mirroring a Latin layout. Type rendering with Arabic ligatures works in the variant pipeline. The visual side has caught up with the language side in a way it hadn't a year ago.

What VIMC01 produces in Arabic now

For a comparable campaign, VIMC01 produces:

  • 5–8× the variant volume in Arabic compared to a traditional agency baseline.
  • 3–4 dialect-and-register variants per audience cohort, where the previous norm was one MSA variant.
  • RTL-native layouts, not mirrored Latin layouts.
  • Inline dialect-classifier scores for every variant, surfaced to the senior reviewer.

The senior Arabic reviewer is still essential — and is now spending their time on the highest-judgement work rather than on the production grind. They steer the dialect mix, audit the classifier disagreements, and approve archetypes. The hours they used to spend writing five MSA headlines they now spend on judgement work that compounds.

What this looks like in performance numbers

The single biggest result we've seen across our Gulf client portfolio: Arabic creative win-rate has converged with English creative win-rate. In 2023, win-rates on Arabic creative were 35–60% lower than on English creative for the same brand. By Q4 2025, the gap had closed to 5–15%. By Q1 2026 it was within statistical noise for half our clients.

Brand-side, this changes the budget allocation conversation. If Arabic-language placements are now performing as well as English ones, the audience-share tilt of the budget should shift to match the audience-share of the market — which in much of the GCC is heavily Arabic-dominant. Most brands we inherit are still allocating based on the legacy creative-volume constraint rather than the current performance data. Reallocating typically finds 8–18% incremental ROAS without spending an additional dirham.

What still needs human eyes

Three categories of Arabic creative still get full human review before publication regardless of how many model layers signed off:

  • Religious and seasonal references. Ramadan, Eid, National Day campaigns. The cost of getting the tone wrong is asymmetric — there's no upside to using a model output without senior review.
  • Comparative or competitive claims in Arabic. Defamation thresholds and the cultural posture toward direct competitor mentions vary across GCC markets. This is judgement work, not pattern-matching work.
  • High-spend launch creative. When an archetype is going to receive seven figures in spend across a quarter, the marginal cost of a thirty-minute senior review is irrelevant compared to the cost of a subtle off-brand miss compounding across a million impressions.

Outside those categories, the volume is high, the win-rate is competitive, and the bottleneck is no longer creative production. It's the same bottleneck as in any AI-native engagement: strategic clarity. Which is, on balance, a better problem to have.