When AI Entered Advertising And Tested The Limits Of Human Connection
2025 pushed generative AI into the center of brand work. Production got faster, asset counts exploded, and visuals went places teams couldn't at this speed or cost.
But the lesson was blunt: audiences didn't reject AI itself. They rejected work that felt empty, uncanny, or misaligned with moments that ask for trust.
What actually happened
- McDonald's Netherlands: A fully AI Christmas film leaned on exaggerated mishaps. Viewers called it unsettling and emotionally flat. It was pulled within days.
- Coca-Cola: An AI-heavy holiday refresh of the iconic trucks traded warmth for morphing spectacle. Fans compared it unfavorably with earlier, human-led spots that felt sincere.
- H&M: Plans to test AI-generated digital twins for social drew pushback from labor groups and creatives. Questions centered on authenticity and the future of paid work.
- Zomato: Its 2024 ban on AI food photos kept surfacing through 2025 as a trust case study. The glossy AI dishes didn't match reality, so menus went back to real images.
- Prime Video (Amazon): AI-generated episode recaps and anime dubs were rolled back after viewers flagged inaccuracies and tonal issues.
- PepsiCo: Announcements about AI in sales and ops drew criticism tied to energy use and sustainability. Even non-creative AI moves can trigger reputational questions.
The pattern marketers can use
Backlash clustered where AI looked like it replaced human judgment in moments loaded with emotion: nostalgia, appetite, identity. People weren't mad that AI was present. They were reacting to the feeling that craft and care were missing.
In short: the audience senses automation. Especially in holiday storytelling, food, and fashion-categories built on humanity.
Practical rules for 2026 planning
- Decide the role of AI per asset. Use it for ideation, versions, resizes, compliance checks, and light comping. Keep flagship films, seasonal tentpoles, and brand-defining imagery human-led.
- Make a "human at the helm" policy. A director, writer, or CD owns emotion and narrative. AI supports, it doesn't drive.
- Run an "uncanny" check in pretesting. Use small audience panels to rate warmth, trust, and coherence. Kill anything that feels eerie or emotionally flat.
- Label synthetic media where appropriate. Stay aligned with guidance and consumer expectations. See the FTC's note on AI marketing claims here.
- Balance the budget. Save with AI on scale work, then reinvest that delta into human craft for the master story.
- Measure by feeling, not just reach. Track sentiment, "warmth" scores, brand lift, and complaint rate by creative origin (human, hybrid, AI-first).
- Plan a rollback path. If a piece lands wrong, have a prewritten statement, fast replacement assets, and updated media plans ready.
- Mind labor and sustainability optics. Engage creators and talent early. Share how you're addressing energy use and data sourcing.
- Protect your IP. Keep sensitive brand assets out of open tools. Log prompts, assets, and model versions for audits.
- Create a data wall for food and fashion. Products that people put on or in their bodies demand realism. Default to real photography and real people.
Signals that say "don't lead with AI"
- Tradition or nostalgia is the anchor (holidays, heritage lines, iconic characters)
- Taste, texture, or fit needs to feel real (food, beauty, apparel)
- Communities with strong creative labor roots are involved (models, illustrators, voice actors)
- Brand trust is fragile after recent news or changes
Signals that say "AI can sit in the background"
- Versioning hundreds of sizes and languages
- Concept boards, colorways, and fast comps for internal alignment
- UGC-style variations with strict QA and disclosure
- Operational content: FAQs, policy pages, CRM snippets with human review
Team moves that pay off
- Stand up an AI creative desk for prompt libraries, model tuning, and compliance guardrails.
- Train producers and brand managers on synthetic media risks, testing protocols, and disclosure norms.
- Codify your "red list" of moments where AI will not lead (by category, season, and channel).
If you need structured upskilling for your team, see this practical certification for marketers here.
How to sell this plan to the business
- Show both sides of the ROI. Present savings from AI at scale and the revenue risk of losing emotional impact on hero work.
- Use case studies. Map the 2025 missteps to specific guardrails you're adopting.
- Forecast by format. AI-heavy for lower-funnel and variations; human-led for brand storytelling. Simple, defensible, and clear.
Trust drives effectiveness. For a broader view on what builds or breaks it, the Edelman Trust Barometer is a useful reference point here.
Bottom line
AI can scale production. It can't fake care.
Use it as an amplifier, not a replacement, especially in moments where people expect heart, memory, and proof. That's where brands win or lose the year.
Your membership also unlocks:
