AI Images, Empty States, and the UX of Starting

Two research findings landed within days of each other this week, and they say something interesting in combination. One: users are paralysed by blank interfaces and equally reluctant to disturb anything that looks finished. Two: AI-generated images now perform on par with stock photography when users don’t know the source. Neither finding is earth-shattering alone. Together, they sketch a practical roadmap for design teams who want to reduce friction without blowing the budget.

The Empty State Is a Conversion Killer You’re Probably Ignoring

Writing on UX Collective, Takuma Kakehi articulates something every product designer has observed but rarely documents: people can’t begin from nothing, and they won’t touch anything that looks complete. The sweet spot is a UI that feels started but not finished — enough scaffolding to give the user a foothold, not so polished that they feel like an intruder.

This has direct revenue implications. In e-commerce flows on Shopee or Lazada, wishlist and cart features that launch empty see significantly lower engagement than those pre-seeded with recommendations or recently viewed items. The blank state reads as abandonment, not opportunity. Grab’s ride-hailing interface has long understood this — the map is never truly empty; it populates nearby drivers before you’ve typed a destination, reducing the cognitive weight of starting a booking.

The implementation fix is straightforward but often skipped in sprint planning: design your empty states last, not first. Treat them as a product moment requiring the same copy, visual, and interaction design attention as your hero flows. For mobile-first SEA markets where screen real estate is compressed, a well-designed empty state with a single clear prompt converts measurably better than a generic “No items yet” placeholder.

AI Imagery Closes the Stock Photo Gap — With Caveats

NN/g’s Rachel Banawa published controlled research showing that when users were unaware of an image’s origin, AI-generated visuals produced no perception penalty compared to stock photography. Task completion rates, trust scores, and aesthetic ratings were statistically equivalent across conditions.

For regional marketing teams, this is genuinely useful. Stock photo libraries skew heavily Western in their representation of people, settings, and cultural context. An AI-generated image of a family sharing a meal in a HDB flat, or a street vendor in a Bangkok market, can be produced on-brand and on-brief in a fraction of the time it takes to commission or license authentic regional photography. Brands like Aeon and Central Group, operating across multiple SEA markets with distinct visual cultures, have real incentive to test this seriously.

The caveat: NN/g’s finding applies when users don’t know. Brand storytelling that leans on authenticity — luxury, healthcare, social impact — should treat AI imagery with more caution. A handbag campaign photographed on a real model in real light carries different brand equity than one that isn’t. The research clears AI for functional UI contexts: hero banners, lifestyle callouts, category imagery. It doesn’t greenlight it for everything.

---

Connecting the Two: The Design System Implications

These two findings point toward the same underlying principle: users respond to signals of liveness and momentum. A partially-populated UI feels like a place where things are happening. An AI-generated image, when well-prompted, can carry the same energy as a real photograph — context, warmth, specificity — without the cost or lead time of a shoot.

For design systems teams, the practical upshot is a content layer rethink. Most design systems specify component behaviour and visual tokens but say nothing about default content states. That gap is where friction hides. A component library that ships with realistic placeholder content — not Lorem Ipsum, not grey boxes — trains product teams to think about the user’s first experience from the moment of composition, not after launch.

For Southeast Asian markets specifically, this matters more than it might in a single-language, single-platform context. A product operating across Thai, Bahasa Indonesia, and Vietnamese needs empty states that work with variable text lengths. An AI imagery workflow needs prompts calibrated to cultural specificity across markets — a default prompt producing generic pan-Asian imagery is barely better than the Western stock photo it replaced. Build the prompt library into the design system, not as an afterthought.

The monetisation angle is blunt: lower onboarding friction increases activation rates. Nielsen Norman research consistently links perceived polish and progress to trust, and trust to purchase intent. If your app’s first-run experience shows users an empty dashboard at a moment when you could show them three useful defaults, that’s a conversion rate problem with a design solution.

The Stakeholder Conversation You’ll Need to Have

Both of these initiatives — better empty states and AI imagery workflows — face the same internal obstacle: they look like small bets. Empty state redesigns don’t make it into board decks. AI image testing sounds like a cost-cutting measure rather than a quality investment.

The framing that tends to work: run an A/B test on a single high-traffic empty state, measure activation rate delta, and express it in revenue terms. A 5% lift in cart activation on a mid-size SEA e-commerce platform can translate to seven-figure annual impact. That’s a design decision worth a sprint.

For AI imagery, the stakeholder concern is usually brand safety, not aesthetics. Address it directly: establish a tiered policy — AI-generated for functional UI and paid social testing, human photography for brand campaigns and influencer contexts. That’s not a compromise; it’s a sensible allocation of creative spend.

The question worth sitting with: if your users can’t tell the difference between AI and stock, and your designers can’t ship fast enough to keep empty states from killing activation — what exactly are you optimising for when you insist on the status quo?

---

At grzzly, we work with digital and e-commerce teams across Southeast Asia to turn design decisions into measurable growth outcomes — from auditing empty state experiences in mobile apps to building AI imagery workflows that respect brand guidelines and cultural context. If either of these threads resonates with a problem your team is sitting on, let’s talk.