A notes app is "notes app" in the US, "not defteri" in Turkey, "ノートアプリ" in Japan, and "app de notas" in Brazil. The query, the spelling habits, the brand abbreviations, the competitors people search for. None of it translates one-to-one. SearchAd AI drafts the keyword list per locale, not per English seed.
Every campaign you approve comes with two ad groups by default. The Exact group carries the long-tail keywords the model drafted for that locale, with EXACT match and Search Match off. The Competitors group carries the rival brand names you told the planner about, drafted with the right spellings, transliterations, and common misspellings for the market.
The split keeps your CPT data clean. Competitor terms inflate cost-per-tap and compress conversion; you do not want them averaging into your generic group. Split them on day one and you save yourself the cleanup on day forty.
You can, and it will quietly underperform. Search habits diverge by market: Brazilians spell apps in Portuguese and English in different proportions, Japanese users mix kanji and katakana, German queries compound nouns into single words. A translation misses all of that. SearchAd AI drafts the list in the locale instead of translating in.
Two sources: the model's understanding of how your app is described in each locale, and competitor brand terms drawn from the apps you mention. Both go into your draft, in two ad groups, so you can act on them separately.
Enough to give the auction real surface to learn from, capped so spend stays predictable. The default is a long-tail Exact group plus a tighter Competitors group. You can ask the planner for more or fewer in chat.
No. Default architecture is Search Match OFF with EXACT match keywords. Search Match makes spend unpredictable and data noisy. We prefer a curated list you can read and adjust.
Try the planner on your own app. The first per-locale list lands in your draft within seconds.