Global coverage

Fourteen at launch: English, German, French, Spanish, Italian, Portuguese, Dutch, Swedish, Danish, Norwegian, Finnish, Polish, Czech and Turkish. Each is a natively authored corpus with its own market context, its own regulatory disclaimer strings and its own documented false friends — not an English list run through a translation service. Chinese, Japanese, Korean, Arabic and Hebrew are not supported at launch, and we would rather say so here than let you find out in a trial. The behavioural Runtime Integrity layer is language-independent and protects every market equally; the keyword packs do not yet.

Because Germany has both the strictest regime and, as a direct consequence, the largest and most sophisticated German-language grey market. The Glücksspielstaatsvertrag 2021 created a licensed online market supervised by the GGL, and licensed operators carry heavy obligations: the OASIS central self-exclusion register, a €1,000 monthly deposit limit, a €1 slot stake cap and the five-second rule between spins. Every one of those constraints generated an evasion vocabulary, and that vocabulary is now the single highest-value detection signal in the German corpus. Casino ohne Lizenz. Ohne OASIS. Ohne Limit. Ohne 5 Sekunden Regel. No licensed operator would ever write those phrases. They are conclusive on sight, and they identify the highest-harm creative a German publisher can serve.

With nuance, because a blanket gambling block is the wrong answer in every Nordic market. Svenska Spel and ATG in Sweden, Norsk Tipping and Norsk Rikstoto in Norway, Danske Spil in Denmark and Veikkaus in Finland are lawful advertisers in their own countries, and many publishers actively want that revenue. So their brand terms are weighted low on purpose — at or below 0.5 — and cannot fire the pack on their own. A publisher who wants them blocked raises the weight in their own policy or adds a Blocklist rule; a publisher who wants them allowed adds one allowlist entry. What the pack is built to catch is the grey market, whose entire SEO vocabulary is built on evading the national self-exclusion registers.

Two distinct obligations, and PubTrust is built to help with both without becoming a problem itself. Under the Digital Services Act, an online platform must ensure that recipients can identify, for each advertisement, that it is an advertisement, on whose behalf it is presented and who paid for it — and must not present advertising based on profiling using special-category data. Very large platforms additionally maintain a public ad repository. PubTrust does not file your DSA disclosures for you, but it gives you the record underneath them: for every creative on your inventory, the advertiser domain the bidder declared, the domain we actually observed, the bidder, the creative identifier, the landing destination and the category — retained, queryable and exportable. Combined with the viewpoint-neutral Political pack, which detects the mechanics of paid electoral advertising so you can label or withhold it, that is the substrate a DSA compliance process needs.

Four markets with four regimes and, in the Portuguese case, two inside one language.

Italy has the strictest advertising regime in the EU. The Decreto Dignità imposed a total ban on gambling advertising and sponsorship of any kind, licensed or not. Essentially every Italian-language gambling creative on publisher inventory is unlawful, which means an Italian publisher has direct legal exposure rather than a brand-safety preference — so the Italian file is weighted slightly firmer than its Romance-language siblings. The false friends are particular: gioco/giochi mean game in every sense, puntata means a TV or podcast episode at least as often as a stake, quota means a share or a membership fee, and casinò with the accent is the gambling house while casino without it is ordinary Italian slang for a mess. Both spellings are carried, weighted differently.

Spain licenses through the DGOJ, and Real Decreto 958/2020 severely restricts advertising — no welcome bonuses to new customers, a 01:00–05:00 broadcast window, no celebrity endorsement — so a large share of Spanish-language gambling creative is non-compliant or offshore. The corpus carries Latin American as well as peninsular vocabulary: tragamonedas for MX, CO, AR, PE, CL alongside Spain's tragaperras. apuesta por is extremely common political and business journalism meaning "backs" or "opts for", so the bare noun is held low and only compounds carry weight. ONCE, the Spanish blind-persons' lottery, is omitted entirely as too dangerous to match.

Portuguese serves two very different markets and both spellings ship. Portugal is licensed by the SRIJ and spells it casino, bónus, apostas desportivas. Brazil legalised sports betting and online gaming under Lei 14.790/2023, licensed from 2025, and spells it cassino, bônus, apostas esportivas. Brazil is currently the most aggressive gambling-advertising market in the entire corpus: the jogo do tigrinho slot craze and the plataforma de apostas influencer wave are Brazil-specific phenomena and carry high weights. Jogo do bicho, the illegal but ubiquitous numbers lottery, is encoded too. Meanwhile jogo/jogos mean game — and in Brazil overwhelmingly a football match — so they sit at 0.15.

France prohibits online casino games in metropolitan France entirely; only sports betting, horse racing and poker are licensed by the ANJ. A French creative advertising casino en ligne is therefore offering an illegal product, which justifies high weights on casino vocabulary specifically. The statutory warning and the 09 74 75 13 13 helpline number are the most reliable strings in the file. Jeu/jeux mean games in every sense including the Olympics, and mise means placement — mise en place, mise en scène, mise à jour — so both are held at 0.15.

The Netherlands and Belgium. The Dutch market opened in October 2021 under the Wet Kansspelen op afstand, regulated by the KSA — and then since 2023 untargeted gambling advertising, explicitly including untargeted online display, has been banned. A Dutch publisher serving a gambling display creative is very likely serving a non-compliant ad, which makes this pack a compliance control rather than a taste preference. Licensed operators must carry Wat kost jou het spel? and point to Loket Kansspel, so those are the strongest Dutch signals. Dutch-speaking Belgium has its own regime under the Kansspelcommissie, which banned gambling advertising almost entirely from 2023; the vocabulary is shared.

Because where gambling is restricted, essentially all of the advertising is grey-market, and grey-market advertising develops its own stable, highly distinctive vocabulary that has no lawful counterpart at all. An English gambling list translated into Turkish would miss nearly everything that actually appears on Turkish inventory.

Turkey. Gambling is illegal except for the state products — İddaa for sports, Milli Piyango for lotteries, TJK for horse racing. There is no licensing route for private online operators, and Law 7258 and Article 228 of the penal code criminalise both operating and advertising online gambling, with BTK blocking tens of thousands of domains. The consequence is a Turkish gambling-ad vocabulary that is entirely grey-market and remarkably stable: deneme bonusu (trial bonus), çevrimsiz bonus (bonus with no wagering requirement), bahis siteleri. No lawful Turkish advertiser writes any of them, so they are weighted as conclusive. Turkish operator domains are deliberately absent from the corpus — BTK-blocked sites reopen within hours under an incremented domain, so a domain list would be stale on arrival; rotating-domain URL patterns do that work instead. Turkish ad copy also strips diacritics routinely, partly to evade filtering, so the highest-value terms ship in both the correct and the ASCII spelling, the ASCII form weighted slightly lower. And the false friends are severe: oyun means game, bahis survives in formal Turkish meaning "topic" (bahsi geçen = "aforementioned"), kupon is an ordinary voucher, tombola is a traditional family game played at New Year, and kar — the de-accented form of kâr, profit — also means snow.

Poland. Online casino is a state monopoly: Total Casino, run by Totalizator Sportowy, is the only lawful online casino, with online sports betting licensed separately to a small set of bookmakers. Poland maintains the Rejestr Domen Zakazanych, a state DNS blocklist, plus payment blocking. So a Polish-language kasyno online affiliate creative is advertising an unlawful product in almost every case. The false friends are brutal: gra/gry/grać/gracz mean game, to play, player; zakład alone means a workplace or a plant at least as often as a bet — Zakład Ubezpieczeń Społecznych, zakład pracy, zakład fryzjerski — so only the compound zakłady bukmacherskie carries weight; kurs/kursy mean a course or an exchange rate far more often than betting odds, so only kursy bukmacherskie counts. Depozyt is ordinary banking and legal Polish.

Czechia. Licensed under the zákon o hazardních hrách, supervised by the Ministerstvo financí, which maintains a public blocklist of unlicensed operators plus payment blocking. The statutory warning — Ministerstvo financí varuje: Účastí na hazardní hře může vzniknout závislost — appears in Czech gambling advertising and essentially nowhere else, making it the single most reliable Czech signal we have. Players must be registered against the national self-exclusion register, so bez registrace positioning is explicit evasion. Sazka, Tipsport, Fortuna and Chance are licensed operators weighted low — and fortuna is additionally an ordinary Czech noun meaning luck, which is exactly the kind of collision that makes a translated list dangerous.

No, and for a different reason than in Europe. US gambling legalisation is state by state: since the fall of PASPA, sports betting and online casino have been legalised on a rolling, inconsistent, still-changing basis, so the same operator can be a licensed advertiser in one state and unlawful in the next. A single national on/off switch is the wrong control. What PubTrust gives a US publisher instead is the ability to see gambling creative by country, region and city — geography is carried on every violation — and to enforce at whatever granularity your own commercial and legal position requires, including Monitor in some places and Block in others. The same logic applies to cannabis, where a licensed dispensary industry advertises lawfully in much of the country and our pack deliberately holds bare category nouns at 0.2–0.35 rather than treating the topic as the violation.

Honestly: partial, and we will not pretend otherwise. Latin America is well covered — Spanish and Portuguese are both first-class corpora with LatAm-specific vocabulary, and Brazil in particular is one of the most heavily weighted markets we ship. India, Singapore, Hong Kong, the Philippines, Australia, New Zealand, South Africa and the Gulf's English-language inventory are covered by the English corpus, which carries UK, US, IE and AU disclaimer vocabulary. Chinese, Japanese, Korean, Arabic, Hebrew, Hindi, Thai, Vietnamese and Indonesian are not supported at launch. For inventory in those languages, the keyword packs will under-read; the always-on Runtime Integrity layer, the pre-render bid metadata layer, domain and URL rules, and the server-side verdict cache are all language-independent and work exactly as well. If CJK or RTL is your core market, tell us in a trial rather than discovering it in production — and if it is on your roadmap, it is on ours.

Every market on this page has a lawful gambling industry that some publishers want, a grey market that almost none do, and a vocabulary in which those two are distinguishable only if somebody bothered to learn it. That is why our corpora carry a written note against every non-English term, why licensed national operators are weighted low on purpose, and why a pack can never block on a single ordinary word.