Why Advertising Got Emotional in 2025

The Ten-Day Problem

It was rainy season on Oahu, the kind where the clouds move sideways and the light can't decide what it wants to do. The rubbish truck clattered up the street, lifting and dropping bins in a rhythm that didn't quite settle. I was at my desk on a Tuesday morning when I noticed something in the ad account I'd been monitoring for three weeks. The creative had stopped working. Not slowly the way campaigns used to decline over months of saturation, but suddenly. Ten days in, the engagement rate dropped by thirty percent. By day fourteen, the client was asking what went wrong.

Nothing had gone wrong, exactly. The ad was the same. The targeting was the same. The budget hadn't changed. What had changed was that people had seen it enough times to predict what came next, and their brains had learned to ignore it.

This started happening everywhere around the same time. Different clients, different industries, all reporting the same compression. Ads that used to run for six weeks were burning out in two. Not because they were bad. Because the pattern had become familiar.

I kept spreadsheets. Tracking when things stopped working and trying to understand why. The numbers kept pointing to the same threshold: somewhere between five and ten days on Meta, three to five on TikTok, three to four weeks on YouTube. After that, you were buying impressions that nobody actually saw. Meta's own performance data showed a forty-one percent drop in click-through rate once people had seen the same ad multiple times. The timeline varied by platform and by industry, but the direction was consistent. E-commerce campaigns were hitting creative fatigue around day seven.

The Systems Running Underneath

Around the same time the creative fatigue accelerated, the platforms rolled out new systems. Meta called theirs Andromeda. It went live in December 2024 with what the engineering team described as a ten-thousand-times increase in processing capacity. The system was built on GPU architecture, which matters because GPUs handle a different kind of math than the processors that ran the old ad systems. They can look at millions of relationships simultaneously. Between a person's recent behavior, the visual composition of an ad, the time of day, the pacing of the video, whether there's a face in the first frame.

The announcement said Andromeda was designed to find "complex latent relationships" between people and ads. Latent means hidden. The system was learning to see patterns that weren't obvious, patterns the old rule-based targeting couldn't identify.

Google had something similar with Performance Max. By May 2024, it was handling eighty-two percent of all ad spend on the platform. Most advertisers weren't choosing it. It had become the default. You gave the system your creative assets and your business goals, and it decided when and where and to whom your ads would appear. The percentage dropped slightly after that as some advertisers rediscovered manual controls, but the adoption remained high. Seventy-five percent in the US, eighty-four percent in the UK.

What both systems had in common was that they no longer relied on what advertisers told them about their audience. They relied on what they could observe about how people actually responded. If a certain kind of creative triggered a certain pattern of interaction (longer view time, a save, a share, the way someone moved their cursor) the system learned to show more of that creative to more people exhibiting early signals of the same pattern.

Meta reported that advertisers using their Advantage+ automation saw a twenty-two percent increase in return on ad spend. The automation handled budget, targeting, and bidding. What you controlled was the creative itself, and the creative had become the signal the system used to figure out who to show your ad to and when.

How Search Works When Someone's Panicking

Nobody called this emotional targeting. The platforms were careful about that. Facebook had gotten into trouble years earlier over leaked research about detecting emotional states, and a spokesperson said explicitly that they don't offer tools to target people based on emotion. Which is true, technically. There's no dropdown menu for "anxious" or "excited."

But the system didn't need labels. It just needed to see what worked.

There's research on how emotional states change the way people search. A study from 2024 tracked search behavior under stress and found that mental state shows up in query language. When people are scared or overwhelmed, their attention narrows. They ask action-oriented questions. They need an answer immediately, and if it's not in the first few results, they're gone.

Google's own research identified six emotional need states that shape search behavior: Surprise me. Thrill me. Impress me. Educate me. Reassure me. Help me. Each one produces different query language. Someone in "reassure me" mode types a question. Someone in "thrill me" mode types two words. The same underlying need (say, finding a restaurant) generates completely different searches depending on the emotional context.

Microsoft research found that while top search results often use less emotive language, users are more likely to click on results that correlate with their emotional state. The search engines weren't explicitly measuring emotion. They were measuring behavior that correlated with emotional patterns, and they were learning which content worked for which patterns.

Maybe this isn't new. People have always searched differently when they're calm versus when they're panicking. What's new is that the systems delivering information have started to optimize for this.

Where Zero-Click Leaves You

Traditional search worked by matching keywords. If you searched for "running shoes," you got pages that contained those words, ranked by how many other sites linked to them and how well the page was structured for search engines. The system didn't particularly care what you meant by the search. It just tried to find the most authoritative page containing those terms.

Semantic search works differently. It converts both your query and every piece of content into mathematical representations (vectors) that capture meaning rather than just words. When you search, the system looks for content whose meaning-vector is closest to your query's meaning-vector. This is why you can ask "what shoes should I buy for marathon training" and get useful results even from pages that never use the phrase "marathon training."

The shift happened fast. By 2025, somewhere between fifty-eight and sixty-three percent of Google searches were zero-click, meaning people got their answer directly from the results page without visiting a site. Voice search had crossed eight billion active assistants worldwide, more than one per person on the planet. Both patterns pushed the same direction: conversational queries that carry more context, answered by systems trying to understand intent rather than just match words.

An industry developed around this. People started calling it Answer Engine Optimization (AEO) to distinguish it from traditional SEO. The principles were different. Not keyword density or link building, but semantic clarity. Structured data. Content that directly addresses the question someone is actually asking, even if they didn't use the exact words you anticipated.

The performance data showed that content optimized for answer engines had a fourteen percent higher click-through rate than traditionally optimized content. Seventy-seven percent of queries ended with an AI-generated response rather than a click to a website. Traditional SEO still drove fifty-three percent of website traffic, but the distribution was changing. The question wasn't whether to optimize for search engines or answer engines. It was how to be visible in both systems at once.

The Numbers on Emotional Content

The performance data supported the shift. A large analysis found that advertising grounded in empathy (ads that reflected the audience's emotional state or values) was seventy-nine percent more likely to drive brand choice than ads focused on product features. Not seven percent. Seventy-nine percent.

Another study across seventeen hundred campaigns showed that emotionally-driven ads had a thirty-one percent success rate in driving profit gains, compared to sixteen percent for rational-only campaigns. That's twice the performance. Ads that triggered above-average emotional response saw a twenty-three percent increase in sales.

Survey data from 2024 showed that eighty-two percent of consumers with high emotional engagement always bought from brands they trusted, compared to thirty-eight percent of those with low emotional engagement. Emotionally connected customers provided three hundred six percent higher lifetime value.

The business case was clear enough that it didn't require theory. Content that acknowledged how someone was feeling outperformed content that didn't, consistently and by large margins. Whether that was because people trusted it more, or because it matched the psychological frame they were in when they searched, or because the AI systems had learned to recognize it as high-performing and gave it preferential distribution, probably all of those things at once.

What You're Actually Controlling Now

By late 2024, most of the manual controls had been removed from advertising platforms. You couldn't really optimize targeting anymore because the machine was doing that. You couldn't really optimize bidding because that was automated too. The main thing you controlled was the creative itself, and the creative had become the signal the system used to figure out who to show your ad to and when.

If your creative carried clear emotional coding (if it was obviously for someone in a specific state, addressing a specific concern) the system could match it to people exhibiting signals of that state. If your creative was generic or emotionally ambiguous, the system had less to work with. It might show your ad, but not at the moments when it would resonate.

This created a strange situation where psychology had become mechanically necessary. Not as a way to manipulate people, but as a way to be legible to the systems that controlled distribution. Content without emotional dimensionality didn't get suppressed, exactly. It just became invisible by default, because the algorithms couldn't figure out who it was for.

The volume requirements changed too. Where you might have run three or four creative variations in 2023, the recommendation by 2025 was eight to twelve active creatives per platform. Not because more was always better, but because the burnout timeline had compressed to the point where you needed a constant rotation. The platforms were testing different combinations of headlines and images and landing pages, learning which emotional frames matched which query contexts. You had to feed the system enough material to run those tests.

One Example That Worked

I had a conversation with a client in October who sold project management software. Their ads had always focused on features. How many integrations they had, how the interface was designed, what the pricing tiers included. The ads were factually accurate. They just weren't working anymore.

We rewrote one to start with a question: "Feeling overwhelmed by scattered tasks and missed deadlines?" The rest of the ad was almost identical. Same product, same features, same offer. But that opening line (acknowledging the emotional reality of someone searching for project management software at midnight) changed how the algorithm distributed it and how people responded to it.

The performance improved by forty percent in the first week.

It wasn't magic. It was just that the ad now matched what was actually happening in someone's head when they went looking for a solution. The old ad had assumed people were calmly comparing features. The new one acknowledged that they were probably stressed, possibly behind on something, looking for help.

The question wasn't difficult to arrive at. Project management software searches spike between ten pm and two am. People searching during those hours aren't doing research for fun. They're behind on something. They're worried about what they forgot. The emotional state isn't subtle.

Other products have different emotional entry points. Someone searching for running shoes at six am is in a different state than someone searching at eleven pm. Someone looking for accounting software in January is in a different frame than someone looking in July. The work is identifying what state your searcher is actually in when they go looking, and then creating content that acknowledges that state without pretending to solve problems you can't solve.

The project management ad didn't promise to fix the deadline that was already missed. It just acknowledged that scattered tasks and missed deadlines were real, and that the person seeing the ad might be dealing with both. That acknowledgment was enough to signal that the product understood the actual problem.

Why People Started Caring About Brand Values

The same pattern was showing up in how people evaluated brands. Survey data from 2024 showed that forty-six percent of Gen Z consumers judged people based on the brands they used. Roughly one-third of US adults said a brand's values were very important in their purchasing decisions. At the same time, fifty-three percent of consumers were experimenting with or regularly using generative AI, up from thirty-eight percent the year before. But trust remained the filter. Consumers who didn't trust a brand wouldn't buy from it, regardless of features or price.

Trust had become the filter. In a market where anyone could generate infinite content, where every competitor could produce professional-looking ads and websites and product descriptions, the question wasn't "is this information available" but "do I believe this source." And that question was answered emotionally, not rationally.

By early 2025, the convergence was obvious if you were watching the numbers. Platform automation had eliminated tactical advantages. Creative fatigue had accelerated to the point where volume alone couldn't solve it. The algorithms running both ad delivery and search results had shifted to optimize based on interaction signals that correlated with emotional resonance, even if they weren't explicitly labeled that way. And consumers were filtering everything through identity and trust because that was the only reliable signal in a world of infinite generated content.

All of this happened in about eighteen months.

Looking at the Patterns

I kept thinking about the client whose ads burned out in ten days. Not because the market had seen enough of the message, but because individual people had seen it enough times to stop noticing. The threshold wasn't about reach or frequency in the traditional sense. It was about the point where the brain could predict what came next and stopped paying attention.

The only way to get past that threshold was to introduce genuine novelty. Not just new footage or new copy, but new psychological entry points. Different emotional frames. Different contexts. The same underlying offer, but approached from angles that didn't feel like variations of something already seen.

This is what the platforms' systems were built to handle. Meta's Andromeda was processing fifteen million ads per month by late 2024, learning which combinations of visual elements and pacing and narrative structure worked for which patterns of user behavior. The volume wasn't the point. The learning was the point.

Google's Performance Max was doing the same thing on the search side, testing different combinations of headlines and images and landing pages, learning which emotional frames matched which query contexts. The systems weren't trying to manipulate anyone. They were trying to solve a matching problem: which message, for which person, at which moment.

What made 2025 different from 2024 wasn't that any single technology launched. It was that all of these systems reached majority adoption at roughly the same time, and the market adapted to them faster than most businesses realized. The old playbook (good targeting, consistent creative, strong offer) still mattered, but it wasn't sufficient anymore. You also needed content that carried psychological signals the algorithms could detect and match to user states they were learning to predict.

I think about this sometimes when I'm looking at a spreadsheet of campaign performance, trying to figure out why one ad worked and another didn't. The numbers show patterns, but the patterns are about people. How they feel when they search, what they need when they see an ad, whether they trust what they're being shown. The machines got better at seeing those patterns. Which means the content has to address them.

Plate Lunch Collective works with businesses trying to figure out what comes next in this shift. The work is mostly about making content legible to these systems without pretending to be something you're not. If that sounds relevant, get in touch.

References

TikTok Paid Media Strategy: What Actually Scales in 2025. Socium Media. https://www.sociummedia.com/blog/tiktok-performance-marketing/

How to Prevent Meta Creative Burnout Before It Kills ROAS. Madgicx. https://madgicx.com/blog/meta-creative-burnout-prevention

Why E-commerce Brands Hit Creative Fatigue in 7 Days. Medium. https://medium.com/@pathwaystoyou/why-e-commerce-brands-hit-creative-fatigue-in-7-days-and-the-volume-solution-nobody-talks-about-85f6b0a9168d

Ad Fatigue Management: When and How to Refresh. DashClicks. https://www.dashclicks.com/blog/ad-fatigue-management-using-white-label-facebook-ads

How Often Should You Refresh Paid Social Creative? Socium Media. https://www.sociummedia.com/blog/avoiding-paid-social-ad-fatigue/

Best Emotional Marketing Statistics 2025. Amra & Elma. https://www.amraandelma.com/emotional-marketing-statistics/

Analysis of Emotion in Marketing. Embryo. https://embryo.com/blog/emotion-in-marketing-stats/

The Psychology of Branding: How Customers Build Emotional Connections. Averi. https://www.averi.ai/blog/the-psychology-of-branding-how-customers-build-emotional-connections

How To Measure Emotional Impact In Branding. Chris Rubin Creative. https://chrisrubincreativ.com/how-to-measure-emotional-impact-in-branding/

The State of Performance Max Campaigns 2025. Smarter Ecommerce. https://smarter-ecommerce.com/blog/en/google-ads/state-of-performance-max-campaigns-2025/

Performance Max Statistics 2025. My Codeless Website. https://mycodelesswebsite.com/performance-max-statistics/

Meta Andromeda: Supercharging Advantage+ Automation. Meta Engineering. https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/

"How" Web Searches Change Under Stress. Scientific Reports. https://www.nature.com/articles/s41598-024-65895-4

The Emotion Profile of Web Search. Microsoft Research. https://www.microsoft.com/en-us/research/wp-content/uploads/2019/06/sp385.pdf

Writing for Confessional Search: The Why, What, and How-To. GPO. https://gpo.com/blog/confessional-search-guide/

How Zero-Click Searches Are Changing SEO In 2025. Jellyfish. https://www.jellyfish.com/en-gb/training/blog/how-zero-click-searches-are-changing-seo-in-2025

A New Era of Search: Trending Towards a Zero-Click Future. Prosek. https://www.prosek.com/unboxed-thoughts/a-new-era-of-search-trending-towards-a-zero-click-future-rise-of-geo-why-google-still-matters/

51 Voice Search Statistics 2025: New Global Trends. Demand Sage. https://www.demandsage.com/voice-search-statistics/

Top 35 Voice Search Statistics You Shouldn't Miss In 2025. Digital Silk. https://www.digitalsilk.com/digital-trends/voice-search-statistics/

A Refreshing Way to Beat Creative Fatigue. LinkedIn. https://www.linkedin.com/pulse/refreshing-way-beat-creative-fatigue-readysetco-fdozc

Meta Ad Creative Testing: Scientific Strategies for 2025. Jos Binitty. https://josbinitty.com/meta-ad-creative-testing-scientific-strategies-for-2025/

AEO vs SEO: Which Strategy Delivers Better Visibility. Nine Peaks. https://ninepeaks.io/aeo-vs-seo-which-strategy-delivers

2025 Connected Consumer: Innovation with Trust. Deloitte. https://www.deloitte.com/us/en/insights/industry/telecommunications/connectivity-mobile-trends-survey.html

Winning with Gen Z: Embracing Intention and Values for Brand Success. Edelman. https://www.edelman.com/trust/2024/trust-barometer/special-report-brand/gen-z-embracing-intention-values-brand-success

Brand Values: 3 Key Insights for Marketers to Look for in 2024. CivicScience. https://civicscience.com/brand-values-3-key-insights-for-marketers-to-look-for-in-2024/

Trust, Attitudes and Use of Artificial Intelligence: A Global Perspective. KPMG. https://kpmg.com/xx/en/our-insights/ai-and-technology/trust-attitudes-and-use-of-ai.html