AI has changed how Southeast Asians find products, but not how they decide: roughly one in four now turn to generative AI tools while shopping, yet recommendations from family and friends still carry more weight than any other influence, with online reviews second. AI has become an added layer on top of commerce, not a replacement for the trust networks underneath it. Marketers who spend their AI budget only on producing more creative are spending against the wrong problem.
Key takeaways
- About one in four Southeast Asian consumers use generative AI tools while shopping, according to impact.com's fourth annual discovery study with Cube and Dentsu, reported in July 2026.
- Family and friends' recommendations still matter most, online reviews rank second and content creators trail both; marketplaces such as Shopee, Lazada and TikTok Shop are the starting point for roughly seven in ten shoppers.
- AI shopping varies widely by market: more than a third of consumers in Vietnam, Indonesia close behind, and Singapore at well under half Vietnam's rate.
- Point AI at understanding first and production second, treat discovery and decision as separate jobs, and localise the behavioural assumption, not just the copy.
- Research-first workflows show up in operating numbers: one Singapore performance agency cut pitch preparation from a week to two days and from three people to one.
Why is nobody budgeting for this gap?
Every AI creative pitch landing in a Southeast Asian marketing inbox this quarter carries the same quiet assumption. If machines now shape how people discover products, machines should now shape how the work gets made. The region's own data tells a different, more uncomfortable story, and marketers who miss it are spending against the wrong problem.
A fourth annual discovery study from impact.com, produced with Cube and Dentsu and reported in late July 2026, found that roughly one in four Southeast Asian consumers now turn to generative AI tools while shopping. That number is real, it is climbing, and it belongs on every planning deck. But when the same research looked at what actually tips a purchase, almost nothing had shifted. Recommendations from family and friends still carry more weight than any other influence in the region. Online reviews rank second. Content creators trail both. Marketplaces such as Shopee, Lazada and TikTok Shop remain the starting point for roughly seven in ten shoppers.
The study's authors frame it precisely: AI has become an added layer sitting on top of commerce, not a replacement for the trust networks underneath it. Read that plainly and the conclusion is simple. AI has changed how people find things. It has left how they decide almost untouched.
What is the discovery-to-decision gap?
Call this space the discovery-to-decision gap, because that gap is where most AI marketing budgets are quietly getting spent this year.
It is an easy gap to miss because both sides of the funnel look healthy on paper. Adoption charts climb. Output volume climbs. Cost per creative asset falls. None of those metrics reveal whether the extra content is reaching a shopper at the exact moment they are open to persuasion, or whether it is simply arriving after they have already texted a friend for advice.
The gap also hides inside an average that flattens huge market differences. In Vietnam, more than a third of consumers now shop with AI assistance. Indonesia sits close behind. Singapore runs at well under half that rate. Any regional creative strategy built around one blended number will end up too aggressive in some markets and far too cautious in others, and the campaign results will never explain why.
What does the gap look like day to day?
Picture a scene playing out across the industry most nights this quarter. A campaign is due in three weeks. A marketing lead is searching for an AI tool that will let a lean team produce more creative without it sounding like nobody actually made it. That search reflects genuine intent, a real question about creative capacity.
What comes back is a flood of platforms competing purely on volume and speed: more variants, more formats, more languages, faster turnaround. So the team buys volume. Four weeks later, thirty versions of an ad are running in a market where three quarters of consumers aren't using AI to shop at all, where the purchase decision is still happening inside a family group chat, and where nobody on the team can point to which of the thirty ads actually spoke to something real. The tool delivered exactly what it promised. The deeper question underneath the original search, how to scale creative output without hollowing it out, was never the question the category chose to answer.
Which three shifts close the gap?
None of what follows argues for using less AI. It argues for using it in a different order.
- Point AI at understanding first, production second. The scarce resource in 2026 isn't creative output capacity. It's a defensible, evidence-backed view of what an audience genuinely cares about and feels tension around, gathered before a single asset gets made.
- Treat discovery and decision as separate jobs. AI-powered discovery and trust-powered decision-making need different assets entirely. One requires visibility and information structured for machines to read. The other requires proof that one human will vouch for to another.
- Localise the behavioural assumption, not just the copy. Translating a single AI-generated concept into six markets isn't localisation when actual AI adoption behaviour differs by a factor of two between those same markets.
Why does Southeast Asia feel this first?
Southeast Asia surfaces this gap faster and more visibly than most regions for a structurally simple reason: the decision layer here is deeply social. Purchases route through family conversations, group chats, creators embedded inside specific communities, and marketplace reviews written by strangers who somehow still feel local and trustworthy. Roughly half of affiliate-driven purchases in the region trace back to trust and validation rather than price or convenience.
That kind of market punishes hollow, mass-produced content faster than markets where buying decisions are made alone. It is also, notably, a market where the loudest AI conversation in adtech and martech right now is almost entirely about production speed. Scan what the industry has published over the past month and the overwhelming share of it covers making things faster: tooling, workflow, education, output volume. Very little of it addresses knowing more before creating anything at all. That imbalance is exactly where the opportunity sits right now.
How do teams close the gap in practice?
This is the model being tested inside Singapore's performance marketing community: run the audience research first, map the tensions, personas and moments a brand can credibly enter, and let the creative brief follow that evidence instead of following render speed.
The proof shows up in operational numbers rather than pitch decks. One Singapore-based performance agency reported cutting pitch preparation time from a full week down to two days, and cutting the team involved from three people to one, while dropping prep cost from around four thousand US dollars to about one thousand. That saving didn't come from generating more slides. It came from removing the manual competitive research that used to sit in front of the actual thinking.
What should you ask before your next tool renewal?
None of this is a case against AI in the creative process. It is a case about sequence, about which question gets answered first.
So before the next tool subscription renews, it is worth sitting with one question. How much of your AI budget is buying you more output, and how much of it is buying you a sharper read on the people who still make up their minds by asking someone they actually trust?
Where does SOMIN fit into the read before the render?
This is precisely the sequencing gap SOMIN is built to close. Rather than starting with output, its analytics stack starts with understanding, pulling together SODA and the SOMONITOR suite, including Brand Tracker, Content Library, Perspective Studies and SoInspire, alongside GWI, to give teams a grounded picture of what audiences across Southeast Asia are actually tense about, market by market, before a single asset gets briefed. For brands trying to close the discovery-to-decision gap rather than widen it, that shift in sequence, research first and render second, is the difference between creative that gets shared inside a trusted group chat and creative that simply adds to the noise.
FAQ
How many Southeast Asian shoppers use AI?
Roughly one in four Southeast Asian consumers now turn to generative AI tools while shopping, according to impact.com's discovery study with Cube and Dentsu. The rate varies widely: more than a third in Vietnam, with Indonesia close behind and Singapore at well under half that rate.
What influences purchase decisions most in Southeast Asia?
Recommendations from family and friends carry the most weight, online reviews rank second and content creators trail both. Marketplaces such as Shopee, Lazada and TikTok Shop remain the starting point for roughly seven in ten shoppers.
What is the discovery-to-decision gap?
It is the space between how AI has changed product discovery and how little it has changed purchase decisions. Budgets that only buy more AI-made output often reach shoppers after they have already asked someone they trust.
Should marketers in Southeast Asia use less AI?
No. The argument is about sequence: use AI to understand audiences, their tensions and the moments a brand can credibly enter first, then produce creative that follows that evidence.
Related reading
- What did the Fable 5 shutdown reveal about Asia's AI platform risk?
- Content localisation in Asia is not translation: three market lessons
- Why global brands need market-specific messaging in the AI age
- APAC and MENA marketing calendar 2026–2027
- Generative engine optimisation: how brands get cited by AI answer engines
