How Often Malaysian Ads Say Halal (2026)

Of 222,921 Malaysian ads running as at their last check, 3,434 carry the word halal in their body copy. That is 1.5%. The specific certification vocabulary is rarer again. 649 ads say "halal certified", 446 name JAKIM, and 108 use "sijil halal". The religious-compliance vocabulary is smaller still: 265 ads say wuduk and 226 say syariah. Explicit exclusion claims barely register, with 52 ads saying "pork free". Set that against what the research says about the market. The strongest available figure on Muslim consumer priorities puts halal ahead of both value and quality as the number one consideration in a purchase. If that is right, and it is the most-cited figure in the field, then the gap between what buyers say matters and what advertisers write in the ad is the finding on this page. There is an obvious counter-explanation and it deserves stating rather than dismissing. Where halal certification is near-universal in a category, saying it may carry no information: an advertiser does not write "electrically safe" on a kettle either. That would predict low mentions in exactly the categories where certification is assumed, and it fits some of the vertical pattern below and not all of it.

Food and beverage carries the highest certification share at 6.2%, then travel at 4.3% and healthcare at 3.9%. Food leading is unsurprising. Food leading at a number that low is not. In the one vertical where halal status is a direct product attribute and a legal one, more than nine ads in ten do not say it. Travel sitting second is the more interesting row, because there halal is not a property of a product but a promise about an itinerary, and the phrase is doing work that a certificate cannot do on its own. Healthcare third fits the same reading: an ingestible is a halal question in a way a sofa is not. The religious-compliance and Muslim-market columns are near zero across every vertical, which tells you those words are not general-market vocabulary at all. They belong to specific product categories, muslimah fashion and cosmetics above all, and those categories are small enough as a share of the whole that they vanish at national scale. A page written for one of them should not read the national figure as evidence that the vocabulary does not matter. Verticals here are model-assigned and multi-label, so an ad can sit in two at once and no column sums to anything.

This page counts words. It does not, and cannot, see whether an advertiser holds halal certification. That boundary is load-bearing rather than decorative. Malaysian law treats an unsubstantiated halal claim seriously, and a page publishing per-vertical halal frequencies could very easily be read as an audit of who is and is not certified. It is not one. An advertiser who says nothing may be fully certified and simply not writing it; an advertiser who says halal may be describing a certificate we cannot verify. Neither reading is available from this data and neither should be drawn from it. There is also a surface limit that pushes every figure down. The probes read ad body copy only. A halal logo on the artwork, which is how a great deal of Malaysian halal signalling is actually done, is invisible here. The obvious second surface is model-generated text recognition over the image, and it was deliberately excluded rather than reported alongside: its coverage cannot be quantified, so publishing it next to a quantifiable figure would invite a comparison neither supports. So the honest summary of the whole page is narrow: in the text of Malaysian ad copy, on one surface, this vocabulary is rare. Whether Malaysian advertising signals halal less than Malaysian buyers want is a question this data can raise and cannot close.

Sixteen phrases across four groups: certification, religious compliance, Muslim-market signalling and exclusion claims. Both Malay and English spellings are probed where both exist, and each is published separately rather than merged, so anyone who disagrees with the grouping can rebuild it from the file. Each figure is an exact-phrase query pinned to the ad body field and counted through a facet distribution rather than an estimate. The index's default matching strategy drops query words until something matches, which would silently turn "sijil halal" into "halal" and inflate the specific probe with the generic one's hits, so the strict strategy is pinned and a validation row for every multi-word probe publishes the count under both. They agree, which is what a phrase probe should look like. Probes are never added together. One ad can say halal and JAKIM and muslimah, and there is no way to compute the union: the field is deliberately not filterable, so no filter-side OR is available. Every figure is a per-probe share and there is no family total anywhere on this page. The denominator is every running Malaysian ad, including categories where the question is meaningless. That is deliberate. The interesting quantity is the share of the market that signals, not the share of a hand-picked subset that plausibly could.

The whole dataset is one CSV under CC BY 4.0, free to reuse including commercially with attribution to AdPlay.ai. It carries every probe with its language and group, its count, its denominator and its share, the per-vertical cut, and the matching-strategy validation rows. Take the figures from the file rather than the page, because the file carries the method note that makes each one interpretable.