What it means when an LLM becomes the gatekeeper of brand discovery in Southeast Asia

IIaasskkeeddtthheessaammeeaassssiissttaanntt,,tthhrreeeeddiiffffeerreennttwwaayyss,,ttooppllaannffoouurrnniigghhttssiinnBBaannggkkookkffoorraaffaammiillyyooffffoouurrwwiitthhaaggeenneerroouussbbuuttuunnssttaatteeddbbuuddggeett..TThheeffiirrssttaannsswweerrggaavveemmeeaaMMaarrrriiootttt,,aaHHiillttoonn,,aannddaarrooooffttooppbbaarrIIhhaaddsseeeennrreeccoommmmeennddeeddiinnrroouugghhllyyffoorrttyyootthheerrppllaacceess..TThheesseeccoonndd,,wwhheennIIssppeecciiffiieedd““bboouuttiiqquuee””aanndd““llooccaall,,””ggaavveemmeetthhrreeeepprrooppeerrttiieessIIrreeccooggnniisseeddffrroommtthheeffiirrssttppaaggeeooffeevveerryyaaggggrreeggaattoorr..TThheetthhiirrdd,,wwhheennIIaasskkeeddoouuttrriigghhttffoorrppllaacceessaaBBaannggkkookkrreessiiddeennttwwoouullddaaccttuuaallllyysseennddaaffrriieennddttoo,,aappoollooggiisseedd,,tthheennrreettuurrnneeddtthheeMMaarrrriiootttt..TThheeccaappaabbiilliittyyiissrreeaall..TThheettaasstteeiissbboorrrroowweedd..

I asked the same assistant, three different ways, to plan four nights in Bangkok for a family of four with a generous but unstated budget. The first answer gave me a Marriott, a Hilton, and a rooftop bar I had seen recommended in roughly forty other places. The second, when I specified “boutique” and “local,” gave me three properties I recognised from the first page of every aggregator. The third, when I asked outright for places a Bangkok resident would actually send a friend to, apologised, then returned the Marriott. The capability is real. The taste is borrowed.

That borrowed taste is now a market force. An APAC Consumer Navigator study of three thousand respondents found that fifty-seven percent use AI to plan travel, that fifty-three percent already notice the output is generic, and that forty-two percent say it nudges them toward familiar, large brands. Read those three numbers together and you have the shape of the thing: a majority delegating the decision, half of them aware the delegate is dull, and four in ten being steered, gently, toward the incumbent.

The assistant picks the most legible option, which is a different thing from the best one.

The assistant picks the most legible option, which is a different thing from the best one. A language model recommending a hotel, a clinic, or a private bank reconstructs the most-cited, most-structured, most-repeated answer in its training and its live retrieval, with quality entering only insofar as the web already encoded it. Whoever has the densest, cleanest, most machine-readable presence wins by default. Google Business profile filled in, TripAdvisor reviews structured, local press that names you, an FAQ page a crawler can parse without guessing. The filter rewards legibility, and legibility runs ahead of merit.

For an operator this reframes a familiar problem. The discovery layer has moved from a search results page, where you could buy your way up, to a synthesised paragraph, where you cannot. An incumbent that can outspend you on advertising cannot necessarily outspend you on being citable. A regional hotel group with eleven properties and patient structured data can appear in the assistant’s answer beside a global chain that spends a hundred times more on media, because the model is summarising the web’s consensus about the category, not auctioning a slot. This is the cheap lever a niche operator has against an incumbent, and most of them have not noticed it exists.¹

An incumbent that can outspend you on advertising cannot necessarily outspend you on being citable.

China is already a chapter ahead, and instructive precisely because it took a different road. On the eleventh of May, Alibaba wired its Qwen assistant directly into Taobao, so a shopper can ask, compare, and buy inside the conversation. ByteDance’s Doubao, with the larger user base, closed the same loop on Douyin weeks earlier. The discovery layer there is a commerce platform that still runs on paid ranking and merchant bidding, a billboard wearing the costume of an oracle. The Western and Southeast Asian assistant quietly validates whoever is most legible. The Chinese super-app assistant openly sells the slot. Both end at the same question for a brand: who decides the shortlist, and on what terms.

The honest disclosure is that none of this is settled. The legibility advantage is real today and may close tomorrow, the moment the assistants start charging for placement the way search did, the way Qwen and Doubao already do. For now there is a window in which being structured and citable beats being large. The operator who treats the machine as the new front desk, and asks plainly what it can read about them, is working a lever the incumbent is too big to bother with. The one who waits will find the summary already written, and their name not in it.

The reflex is to call this a marketing task and hand it to the agency that bought your search ads. It is closer to a librarian’s task, and the people who win it tend to be the ones who treat their own web presence as a record to be catalogued rather than a billboard to be rented. ↩

Footnotes