A customer asks AI for a shortlist, then checks the price, reviews and website before getting in touch. For Malaysian businesses, that second step can matter just as much as the AI answer itself.
This is one possible path, not a rule for every buyer. Still, new research on AI shopping behaviour suggests that an AI answer often starts a closer look rather than ending one. The useful question for a business is not only whether it appears in an answer. It is also what someone finds when they check.
The answer is not the end of the buying journey
Think of a fictional customer in Shah Alam looking for an office printer. She asks an AI tool for a shortlist of models that can handle colour printing for a small team. One answer points her towards a local supplier. She opens its website, checks the price and looks for delivery details.
The site says “contact us for pricing”. A marketplace listing shows a figure, but it is unclear whether toner is included. A review mentions slow support, while the supplier’s page says little about repairs. She may still call, but now she has questions that the shortlist did not settle.
This example is made up. It shows a common kind of decision, not a measured Malaysian buying journey. The point is that the next page, listing or review can shape what a buyer does after an AI reply. That makes the handoff from discovery to checking worth a close look.
A business can have a strong product and still leave basic facts hard to find. A buyer might not know whether the price is current, whether an item is in stock or whether a service covers their area. Those gaps create extra work for the person trying to buy.
Our guide to AI discovery in Malaysia looks at the wider set of places where that journey can begin. Here, the focus is narrower: what happens when a person chooses to check an AI suggestion against other sources.
What AI shopping behaviour in Southeast Asia reveals
The e-Conomy SEA 2026 report offers a useful starting point. Its Google/YouGov research covered 9,006 AI users across six Southeast Asian markets in July 2026. The figures describe that group; they are not estimates for all Malaysians.
On page 30, the report says 69% use AI to compare product options. That makes comparison the leading shopping use case shown in its chart. A buyer might ask for a side-by-side view of features or help narrowing a long list. Neither action means they have chosen a seller. The survey captures AI shopping behaviour among AI users, not the whole population.
Page 31 looks at what happens next. More than 90% of users conduct further research or consult reviews after an AI recommendation. The report lists online searches, shopping-platform reviews and social content among the follow-up actions. Its question-level subgroup details are not fully disclosed, so the result should not be treated as a precise rate for every shopper or every recommendation.
The report sums up the pattern this way:
“While 93% of SEA users trust AI as a shopper assistant, almost all continue with further research to validate recommendations.”
In other words, trust and checking can sit together.
Google’s analysis of the report also points to research after an AI answer. For a Malaysian firm, the regional findings are a reason to inspect the next step. They are not proof that a particular share of its own customers follows that path.
Trust does not mean blind acceptance
The 93% figure needs its full context. On page 31, 50% of surveyed AI users said they trust AI but will verify its recommendations. Another 29% trust it only for basic information. The report also shows smaller groups who trust AI without verification or to take action, and 7% who do not trust it at all.
So “93% trust AI” does not mean 93% will accept an answer without question. It includes people who put limits on that trust. For a business, the distinction matters: a confident-sounding answer cannot replace clear facts that a customer can inspect.
The same caution applies to future purchases made by AI. On page 67, 75% of the surveyed AI users said they would be comfortable letting AI find the best price and buy a product for them. The question explicitly asked them to assume the assistant was generally reliable. This is hypothetical comfort under that condition, not a measure of purchases already made by AI.
Separate Adyen research in Malaysia found that 52% of 1,031 surveyed adults were uncomfortable letting AI complete purchases for them. It asks a related question, but it does not use the same population or reliability assumption. The two percentages should not be compared as if they came from one poll.
For now, a practical reading of AI shopping behaviour is simple: make room for both interest and caution. Someone may welcome help with a shortlist while still wanting control over the final decision. Our article on agentic commerce in Malaysia explores the broader shift towards AI acting on a buyer’s behalf.
Find the verification gap
A verification gap is the space between what a buyer needs to check and what they can easily find. It is not a score or a claim about an AI tool’s ranking system. It is a practical way to review the journey from the buyer’s side.
Follow one buyer’s questions
Return to the fictional printer buyer. Start with the question she might ask: “Which colour printer suits a small office?” If your business sells one, note the claims that would matter to her: running costs, warranty, delivery and support. Then try to confirm each claim without relying on an AI summary.
Open the product page on a phone. Follow the link from a listing to the seller’s site. Check the details that appear on each page. If the model name, price or warranty differs, find out which is current. A mismatch may have a fair explanation, but the buyer should not have to guess it.
Do the same for a service. Imagine a café owner in Johor Bahru looking for a company to install a point-of-sale system. An AI reply names a provider, but the owner still needs to know whether the company serves Johor, what setup involves and how to get a quote. This is another fictional example, not a claim about that provider or market.
Record what is missing
Make a short list of the questions a real customer might ask before contact. Next to each one, record where the answer appears and who keeps it current. Include your own website, relevant listings and any public review profiles you manage.
Look for gaps, not just missing pages. A price can exist but apply only to a different package. A delivery promise can omit the areas it covers. A case study can mention a result without explaining the work or the time period. An answer may be technically present yet still be hard to use.
AI shopping behaviour can vary with the buyer and the purchase. Do not assume every buyer follows the same route. Ask your sales or support team which questions come up often, and review any feedback you already have permission to use. These are clues for your own business, not proof that the regional survey applies to every customer.
Make the next step easy to check
Responding to AI shopping behaviour does not require a long list of tricks. Start with the information a person needs to make a sound choice. Then make sure there is a clear route to ask about anything that cannot be settled online.
Put the practical facts in reach
Price and scope: Show a current price where you can. If a quote depends on the job, say what affects it and what the buyer should share. For a package, explain what is included and flag common extra costs. Avoid presenting an old promotion as today’s price.
Availability and coverage: State whether an item is in stock when that information is reliable. For services, explain the locations you cover, expected lead times and any limits. If stock or schedules change quickly, say how a buyer can confirm them.
Evidence: Support specific claims with useful detail. A genuine case study might explain the client’s problem, the work done and the result, with permission to name the client. If you cannot share a name, say so and avoid implying that an anonymous example is independently verified.
Reviews: Make it easy to find genuine third-party feedback where it exists. Read recent criticism as well as praise. Respond accurately to issues you can address, without buying reviews or suggesting that every customer had the same experience. A review is another person’s account, not a guarantee of your next buyer’s result.
Check the handoff, then keep it current
Contact: Show a working phone number, email address or WhatsApp link, along with any useful business hours. Test the link on a phone and confirm that it reaches the right team. If you offer a form, ask for only the details needed to take the next step.
Give someone ownership of these checks. A monthly review may suit a stable service page; stock and prices may need attention more often. When a detail changes, update the relevant pages and listings you control. Keep a note of the source and date so the next person can check it.
Basic access to information matters too. Google’s guidance for AI features in Search says a page must be indexed and eligible to appear with a Search snippet to be eligible as a supporting link in AI Overviews or AI Mode. Meeting those conditions does not guarantee inclusion. Google says no special schema is required for these features.
That guidance is about Google Search features, not a promise about every AI assistant. Clear, accurate pages help people check your business whether or not an AI system cites them. For a wider discussion of being considered in the first place, see the AI recommendation gap. For ongoing search work, explore CODE/RAVEN’s Rank General protocol.
Frequently asked questions
Do customers always check an AI recommendation?
No. The regional research found that more than 90% of users conduct further research or consult reviews after a recommendation, but it does not establish that every customer does so. The question-level subgroup is not fully disclosed, and the survey covered AI users rather than all shoppers.
Does the 93% trust figure apply to all Malaysians?
No. It comes from a survey of 9,006 AI users across six Southeast Asian markets. The figure includes people who trust AI but verify its advice, as well as those who trust it only for basic information. It is not a Malaysia-wide population rate.
What should a small business check first?
Start with one important product or service. Check its price or quote process, availability, supporting evidence, third-party reviews and contact route. Fix conflicting or outdated details that a buyer could encounter while comparing options.
Does a business need special schema to appear in Google’s AI features?
Google says no special schema is required for AI Overviews or AI Mode. A page must be indexed and eligible to appear with a Search snippet to be eligible as a supporting link, but that does not guarantee it will appear.
Should a business let AI complete purchases for customers?
That depends on the business, its customers and the controls available. The regional 75% figure measures hypothetical comfort under an assumption of generally reliable AI, not actual adoption. Start by understanding where customers want help and where they want to keep control.
If you would like a clearer view of the journey after an AI answer, contact Xwork for a conversation about what your customers can find and check.
