AI Is the New Friend Who Knows Everything

Most of us in Singapore still Google things. That hasn’t changed.

But the way we search is quietly shifting. It used to be a keyword. “Best pasta Tanjong Pagar.” Now it’s a conversation. “I’m going on a first date. What’s a good restaurant in Tanjong Pagar or Raffles Place? Somewhere with good pizza and pasta, not too loud?” And instead of typing that into Google, more people are bouncing it off Gemini, ChatGPT, or your new favourite LLM.

Not because we’ve abandoned Google. But because that kind of question (contextual, layered, specific) feels more practical and natural to ask an AI than to key into a search bar and sift through lists and articles.

That’s the shift worth paying attention to. Not Google dying, but a new type of query emerging. And as our trust in AI responses grows — as we find the answers useful and good enough to act on — these AI-assisted decisions are going to happen more often, and at higher stakes.

The research gets easier. The purchase follows faster. And the businesses that get recommended in that first AI response? They’ve already won half the battle.

A new kind of query

AI assistants have become a genuine first stop for recommendation-style searches — the kind that used to require a WhatsApp poll or twenty minutes on TripAdvisor, TimeOut, HungryGoWhere.

“Where should I eat before the show?” “Plan me a 5-day Beijing trip — I like food, culture, and running.” “Best running shoes for wide feet under $150.”

In fact, you would have noticed that AI has already changed the way search engines work. According to McKinsey, About 50 percent of Google searches already have AI summaries, a figure expected to rise to more than 75 percent by 2028”.

These aren’t edge cases. They’re the queries that used to feed an entire ecosystem of review platforms and content sites. What’s changed is that AI handles the synthesis step — it collects, filters, and presents a confident answer so you don’t have to. Whether you act on it is still your call. But fewer people are bothering to dig further themselves now.

What’s actually happening when the AI answers you

Here’s the thing that matters: the AI doesn’t know the best pasta place in Tanjong Pagar. It doesn’t have taste buds. It hasn’t eaten there.

What it does is synthesise from what’s available online — and more specifically, from what’s clearly structured, specific, and credible enough for it to extract and relay confidently.

The restaurant that gets recommended isn’t necessarily the best one. It’s the one whose digital presence is most readable to the model. A spot with a clear description, consistent mentions across multiple sources, and specific attributes (late-night kitchen, great for solo diners, not too loud) is much easier for AI to confidently surface than a hidden gem with a sparse Google listing and no other digital trail.

The Beijing itinerary example is a good one. Ask ChatGPT to plan five days for someone who’s into food, culture, and running, and you’ll probably get something genuinely useful — a solid mix of neighbourhoods, food stops, and routes that would hold up in real life. I’ve had the same experience asking for a 5KM running route in Tokyo.

But here’s the nuance: what you get is shaped by what’s been written about clearly online. A neighbourhood that’s perfect for a running-focused traveller but has never been framed that way in any blog or guide is unlikely to surface — not because it’s a bad suggestion, but because the AI has no confident basis to make it. It’s a reflection of the content landscape, not a comprehensive survey of the actual world.

What it means if you’re a business

Three scenarios are playing out right now across every consumer industry:

The great-but-invisible business. A restaurant with outstanding food, loyal regulars, and zero structured online presence doesn’t get cited. Not because it’s bad — because the AI has nothing to confidently pull from. No attributes, no consistent descriptions, no mentions across sources it trusts.

The well-labelled winner. A hotel that describes itself as “solo female traveller-friendly,” with content across its own site and third-party listings that consistently uses that framing, starts surfacing every time someone asks for that type of stay. The alignment between what the business offers and how AI queries are phrased is doing real work.

The legibility play. Some businesses aren’t necessarily the best option, but they’re the clearest one. Their content answers the specific question being asked. They show up not by being exceptional, but by being the most straightforward answer to a well-defined query.

None of this is cynical — it’s just how the new discovery layer works. It’s worth knowing.

How this is different from the B2B version

GEO (Generative Engine Optimisation), the practice of making content more visible and citable in AI search responses, gets talked about a lot in B2B contexts.

B2B GEO is about influencing a long research cycle. Multiple decision-makers, multiple touchpoints, weeks of consideration.

B2C GEO is faster and more immediate. The moment of intent is short. Someone asks, gets an answer, and often acts. If your restaurant, hotel, or product isn’t in that answer — the decision may already be made. Different stakes, but the underlying idea is the same: be the clearest answer to the question being asked.

Getting more out of AI recommendations

The most useful thing you can do as a consumer isn’t to distrust AI recommendations — it’s to get better at asking.

A vague query gets a vague answer. “Best restaurant in Tanjong Pagar” will surface the same five names every time. But the more context you give, the more useful the response becomes. Tell it who you’re going with, what kind of atmosphere you’re after, dietary restrictions, budget — and you’ll start getting answers that are actually tailored to your situation rather than a generic crowd-pleaser.

If an AI recommendation doesn’t feel quite right, the fix is usually more context, not a different tool. Think of it less like a search engine and more like asking a well-read friend who needs a bit of background before they can give you a good answer. The more you give them, the better they’ll do.

The honest take

AI recommendations aren’t bad. For plenty of queries they’re genuinely useful — faster and more contextual than scrolling through star ratings.

But they’re shaped. They reflect the content ecosystem, not ground truth. The gap between “most recommended by AI” and “actually the best” is real, and it’s growing as more businesses figure this out.

Knowing that makes you a more intentional user of these tools. And if you’re building or marketing anything consumer-facing, it’s probably the most useful reframe you can apply to your content strategy right now.

Go and ask ChatGPT for that Beijing itinerary — the one optimised for food, culture, and running. See what comes up. Then ask the same question about your favourite hawker stall.