Somewhere in the last year, the job of running Google and Meta ads quietly changed. Manual bid adjustments, granular audience layering, endless A/B split tests — the levers that used to define "good ad management" are being absorbed into the platforms' own AI. For brands in Singapore and Malaysia, the practical question isn't whether to let the algorithm drive. It already is. The question is what you still need to control yourself.

Google Has Moved Past Performance Max Into AI Max

Performance Max normalised handing Google's algorithm control over placement and bidding. In 2026, AI Max pushes further — generating and testing variations of your headlines, descriptions, and images automatically alongside your original creative, with AI-generated assets flagged in reporting so you can see what the algorithm is running without your sign-off. The direction is unambiguous: less manual lever-pulling, more automated decisioning built on the signals and assets you feed it.

The catch is what this shifts the actual work onto. Accounts with clean conversion tracking and accurate primary conversions are outperforming accounts with bigger budgets but messy data — the algorithm can only optimise toward what it can measure correctly. For SG/MY brands still running loose or duplicate conversion events, that's the fix that pays off before any creative or targeting change does.

Meta's Advantage+ Is Already Outperforming Manual — With a Catch

The data here is hard to argue with: Advantage+ Shopping campaigns are running at roughly 32% lower cost per acquisition than manually configured campaigns, and advertisers report about 22% higher ROAS on average with AI-driven campaign types. In Singapore specifically, AI-driven formats now outperform manual campaigns on most accounts spending under S$50,000 a month — which covers the large majority of SME advertisers in the region.

The nuance is in the creative supply, not the targeting. Meta's algorithm performs better with more creative variation, not cheaper variation — accounts feeding it a wide, genuinely different set of assets outperform those uploading minor edits of the same ad. Roughly 90% of advertisers globally now use generative AI somewhere in their creative workflow, but most are using it to cut production cost rather than increase the range of what they test. That's the gap worth closing: use AI to produce more distinct creative directions, not ten versions of the same one.

One more thing worth flagging for higher-consideration categories — AI-generated creative shows a modest CTR advantage on Meta overall, but that edge reverses on higher-ticket purchases, where it converts somewhat worse than human-made creative. If you're selling something considered, don't assume AI-generated assets are automatically the better performer.

The Highest-Converting Format in Malaysia Isn't What You'd Guess

While the ad platforms automate targeting and bidding, the format doing the real work in Malaysia is refreshingly simple: Click-to-WhatsApp ads are now the highest-converting format on Meta for service businesses in the market. It's not a trend so much as ads finally matching how Malaysians already prefer to communicate and buy — a direct chat, not a landing page form. If you run a service business and haven't tested this format against your standard traffic or lead-gen campaigns, it's a fast, low-risk experiment to run this quarter.

Content: Short-Form Video Is Still Where Attention Sits

The ad automation shift doesn't change what content actually earns attention in this region — it just means the algorithms have gotten better at distributing whatever you feed them. Short-form vertical video continues to generate roughly 2.4x higher engagement than static content among Malaysian audiences, and Singapore users now spend more daily time on short-form video than on scrolling static feeds. Singaporeans spend an average of 34 hours a month on TikTok alone — more than any other app.

The platform roles are becoming more distinct rather than interchangeable: TikTok for discovery, Instagram for identity and brand, YouTube for longer-form trust-building, WhatsApp and Telegram for the conversion conversations happening in private, not public feeds. Treating all platforms as the same content dumped in different aspect ratios is increasingly the wrong move — the brands doing well are building for each platform's actual job in the funnel.

What To Actually Do With This

Fix your conversion tracking before you touch bids or creative. Every AI-driven system — PMax, AI Max, Advantage+ — is only as good as what it's optimising toward. This is the highest-leverage, least glamorous fix available right now.

Feed the algorithm range, not repetition. Ten meaningfully different creative concepts will outperform ten small edits of one concept, on both Google and Meta.

Test Click-to-WhatsApp if you're a service business in Malaysia. The conversion data on this format is strong enough that not testing it is the riskier choice.

Match content to platform intent. Discovery content for TikTok, trust content for YouTube, identity content for Instagram — don't force one asset to do every job.

Keep a human on strategy and judgment calls. Let AI handle bidding, placement, and creative testing volume. Keep positioning, offer, and what the brand actually stands for firmly in human hands — that's the part the algorithm still can't do for you.

Conclusion

The platforms aren't asking permission anymore — Google and Meta have both made automation the default path, and the accounts seeing the best results are the ones that stopped fighting it and started feeding it well: clean data, genuinely varied creative, and content built for what each platform actually does in the customer's journey. If you want a clear-eyed look at where your ad accounts and content strategy currently stand against this shift, YEA Business works with brands across Singapore and Malaysia on exactly this.