Emotion Data and AI Trust: Fixing Your Ad Stack in 2026

Brands are spending more on creative testing than ever — and most of it is still measuring the wrong thing. Meanwhile, the AI systems quietly running their ad stacks are eroding the customer trust those same campaigns are trying to build.

Emotion Data Is Rewriting the Creative Brief

Click-through rates tell you what people did. Emotion data tells you why they almost didn’t. As MarTech reports, biometric and facial-coding tools can now map audience emotional responses to specific creative frames — identifying the exact moment a spot loses attention or triggers anxiety — before a single dollar goes into media.

For programmatic buyers, this is a material shift. When you’re running DV360 or The Trade Desk across SEA markets with multilingual creative variants, the cost of a misreading between the Thai-language and Bahasa Indonesia cuts isn’t just creative waste — it’s bid efficiency destruction. A 15-second pre-roll that triggers discomfort in the first three seconds will tank your VCR, which feeds back into your auction signals, which quietly degrades your CPM competitiveness over time.

Practically, this means integrating emotion testing into the pre-launch phase — not as a replacement for A/B testing, but as a diagnostic layer that identifies which creative hypotheses are even worth testing at scale. Platforms like Realeyes and Affectiva have made this accessible below the enterprise threshold. The benchmark question to ask: does your current creative review process catch emotional friction before media activation, or after?

The AI Trust Gap Is a Media Efficiency Problem

New research cited by MarTech puts a sharp number on something most performance marketers have been sensing: consumers are using AI tools for product discovery and purchase decisions in significant numbers, but trust in those tools remains structurally low. Usage is ahead of trust — and that gap is where your retargeting assumptions start to break.

In SEA, where platforms like Shopee and Lazada have built AI-powered recommendation engines directly into the purchase funnel, this trust deficit has specific consequences. A shopper who receives a product recommendation they perceive as intrusive or inaccurate doesn’t just ignore it — they develop attribution skepticism. They complete the purchase through a different path, and your last-click model misreads the signal entirely.

For paid media teams, the implication is less about adjusting creative messaging and more about auditing where AI-automated decisions sit in your customer journey. If your DSP’s predictive audience segments are feeding into your CRM’s triggered email flows without a human review layer at any point, you’ve built a trust-erosion machine that optimises toward conversion while quietly destroying retention.

---

When Automation Creates Friction You Can’t See in the Dashboard

MarTech contributor Alicia Arnold makes a point that deserves more airtime in media planning conversations: AI systems misread signals, and when they do, the friction customers experience is real even when your reporting shows green. An automated decision — a suppressed ad, a mis-timed offer, a wrongly excluded audience segment — doesn’t generate an error report. It generates silence, and silence looks like organic behaviour until your cohort retention numbers tell a different story three months later.

This is particularly acute in markets like Indonesia and Vietnam, where first-party data infrastructure is still maturing and many brands are relying on modelled audiences rather than observed behaviour. The error rate in those models is higher, the feedback loops are longer, and the automated decisions downstream are operating on shakier foundations than the platform UI suggests.

The tactical fix isn’t to dial back automation — it’s to build explicit checkpoints. Quarterly audits of which audience segments your DSP has systematically suppressed. Regular cross-referencing of AI-triggered CRM actions against customer service complaint data. These aren’t glamorous, but they’re the difference between a stack that compounds signal and one that compounds error.

Data Silos Are Still the Root Cause — Culture More Than Technology

The March MarTech Conference surfaced a finding that should reframe how marketing and tech leadership prioritise investment: the primary barrier to unified customer insights isn’t the data infrastructure — it’s the operating model. Teams working in silos produce siloed data, regardless of how sophisticated the CDP sitting underneath them is.

For paid media and programmatic specifically, this shows up as a coordination failure between the team managing upper-funnel brand spend and the team managing lower-funnel performance. They’re often reading different dashboards, optimising toward different KPIs, and inadvertently cannibalising each other’s signals. A brand awareness campaign that successfully shifts consideration shows up as a conversion-rate improvement in the performance team’s model — but the attribution never connects, so the brand investment looks inefficient and gets cut.

The fix requires deliberate structural change: shared KPI frameworks, joint weekly reviews between brand and performance leads, and a single agreed-upon measurement methodology. Several regional brands — including some in the FMCG and telco sectors across SEA — have moved toward unified growth teams that hold both awareness and conversion accountable to a single revenue outcome. The technology follows the culture, not the other way around.

---

Key Takeaways

• Run emotion data diagnostics on creative before media activation — identify emotional friction points before they degrade bid efficiency and VCR at scale.
• Audit every AI-automated decision point in your customer journey quarterly; silence in your dashboard is not the same as absence of friction.
• Unifying customer data requires resolving operating model and cultural misalignment first — no CDP purchase fixes a siloed team structure.

The ad stack in 2026 is more capable than it’s ever been — and more capable of optimising confidently in the wrong direction. As AI automation deepens across every layer of the funnel, the question worth sitting with is this: are your measurement frameworks actually designed to catch what your systems are getting wrong, or are they only built to confirm what they’re getting right?