AI & Marketing Research with Dr. Eva Wolf
Evidence-led briefings that translate peer-reviewed studies and important preprints on artificial intelligence, generative AI, marketing, advertising, consumer behavior, and business strategy into practical insight. Dr. Eva Wolf explains what the evidence actually says, what deserves a deeper read, and what marketers, consultants, educators, and business leaders can do next.
AI & Marketing Research with Dr. Eva Wolf
AI Attribution, Trust & Personalization: 3 Marketing Research Signals
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Thanks for listening to AI & Marketing Research Radar by Big Plans Media.
I’m Dr. Eva Wolf, and I help marketers, educators, consultants, and business owners turn AI marketing research into practical strategy, smarter workflows, and real business opportunities.
More episodes: https://bigplans.media/ai-marketing-research-radar/
Consulting: https://bigplans.media/ai-marketing-consulting/
Big Plans Media — Where Big Ideas Meet Smart Marketing.
You're listening to AVIT, an AI-generated research briefing avatar trained on the research framework and methodology of Dr. Eva Wolfe, marketing professor, AI researcher, and founder of Big Plans Media. Every day, Evita scans emerging research in AI, marketing, consumer behavior, psychographics, and business strategy to identify the most relevant developments, opportunities, and risks worth watching. These daily radar reports are designed to help busy professionals stay informed without having to read hundreds of research papers themselves. And every Friday, join Dr. Eva Wolfe live for her personally recorded weekly AI Marketing Radar Roundup, where she breaks down the biggest stories, explains what actually matters, and shares practical insights and strategic implications for marketers, educators, entrepreneurs, and business leaders. Now, here's today's radar report. Here's the signal I can't ignore today. You're running AI across your marketing stack. Recommendations, chatbots, attribution models, and you're assuming it's working. But what if your attribution is missing one in seven meaningful customer interactions? And what if the AI touch points you're counting on to convert are quietly destroying the trust that was supposed to get you there. Today's papers point to the same pattern. Deploying AI in marketing isn't enough. How you deploy it and what you actually measure determines whether it converts. We screened 380 papers. Three cleared the full text bar and made the radar. Quick caveat: this is a first pass research briefing, not a final academic review. Every paper covered today has full text access. I'll tell you what the papers suggest, what they don't prove, and which ones deserve a deeper read. Okay, let's get into it. Paper one. Here's the business question. Is your conversion attribution model lying to you? Think about this. A customer browses cameras on your site. Three days later they buy a memory card. Did that browsing session influence the purchase? Of course it did, but your attribution model probably didn't count it. Because the two products aren't in the same category, and there's no obvious shared purchase history linking them. That's exactly what this paper tackled. The researchers out of Alibaba built a benchmark called Silva, 1,000 conversion events, nearly 70,000 preceding touch points from Taobao. Human annotators labeled each one irrelevant, explicitly related, or implicitly related. Then they ran multiple large language models against it to see if AI could find what rule-based systems miss. Here's what they found. Almost 15% of touch points that led to actual purchases were what the researchers call hidden connections. Cross-category browsing that current attribution systems completely ignore. Hmm, so let me translate that. If you're running attribution on meaningful e-commerce traffic, roughly one in seven significant customer interactions is getting zero credit. Zero. That credit is probably going to the last click or to a direct session. While the browsing behavior that actually primed the purchase is invisible to your model. That is not a data quality problem. That is a budget allation problem. Now here's the catch. The AI models that performed best, flagship models, top tier, were good on the obvious touch points and decent on the hidden ones, but there's still meaningful room to improve on implicit connections. And smaller models, under 100 billion parameters, they lagged badly. So this isn't a plug in any LLM and go situation. Yeah, that's the part I keep coming back to. The gap isn't between AI and no AI. It's between which AI you use and how you prompt it. One more finding worth flagging. Asking the AI to compare two touch points head to head, pairwise, worked better than showing it a whole list and asking it to rank. Though that gap closes with the most capable models. The researchers also took the AI-identified touch points and used them to retrain a purchase prediction model. That model outperformed the production baseline by 0.35 percentage points on their accuracy metric. 0.35 sounds tiny. At hundreds of millions of daily active users, it is not tiny. Plain English payoff. Your attribution model is probably ignoring about 15% of the browsing behavior that actually drives purchases, specifically the cross-category stuff. And AI can find it, but only if you're using a flagship model and prompting it correctly. Okay, here's where this becomes commercially interesting. Money move. Build or buy a semantic touch point audit. Run your historical conversion data through a large language model. Surface which cross-category customer journeys your attribution model is currently ignoring, and show your team exactly where credit is being misallocated. That's a consulting engagement, a SaaS feature, or an internal data science sprint, depending on your setup. Action step. Before your next campaign review, ask your data team one question. How does our attribution model handle cross-category browsing? If the answer is it doesn't, you have a gap worth quantifying. Evidence check. This is a preprint accepted to CIKM26, but not fully through peer review. More importantly, the benchmark is built entirely from one platform, Alibaba's Taubao. If you're not running a massive Asian e-commerce operation, treat the specific numbers as directional, not literal. Radar verdict, deep dive. The methodology here is unusually rigorous for a preprint. Nearly 70,000 labeled touch points, substantial interannotator agreement, multi-model evaluation, and downstream validation on an actual conversion model. If you work in e-commerce attribution at scale, read the full paper. This next one looks like a trust study. But stay with me because the business implication is bigger than it appears, especially if you're running AI automation in any customer-facing journey. Paper two. This research surveyed 438 consumers who had already experienced AI-enabled digital marketing. They looked at three things. Did human AI collaboration increase trust? Did it increase purchase intent? And did it affect how confidently people made financial decisions? They used a structural equation model to map the relationships between those variables. Here's what they found. When consumers perceived that a human was involved alongside the AI, trust went up. And trust was the strongest effect in the entire model. It then flowed into both purchase intent and financial decision confidence. Huh. The model explained 64% of the variance in purchase intention. That's not nothing. So let me translate that. If your brand is running AI chatbots, AI recommendations, or AI generated content, and there's no visible human involvement, you're leaving trust on the table. And in this model, trust is the engine that drives purchases. Not more AI, more visible human accountability. Here's the catch. This is a cross-sectional survey. Everyone answered at one point in time, so we can't confirm that human AI collaboration causes more trust, only that the two are associated. And the sample is 438 consumers, likely in India, who already had prior experience with AI marketing. People unfamiliar with AI may respond completely differently. This actually bothers me. The study treats human AI collaboration as one composite variable. It doesn't tell us which specific behaviors drive the trust. Is it a disclosure statement? A human photo next to the AI recommendation? A named advisor? We don't know. And that matters enormously for implementation. Plain English payoff. Consumers trust AI marketing more and buy more when they can see a human was involved in the process. If you're running fully automated AI customer journeys, adding any visible human accountability signal is worth testing. Okay, here's the business hiding inside the research. Money move. Productize a human in the loop overlay for AI-assisted e-commerce or fintech experiences. Think AI recommendations badged with reviewed by our team or a named specialist visible in the flow. For financial services specifically, this is a premium conversion feature, not a cost. Pitch it that way. Action step pick one AI-assisted touch point in your customer journey: a chat bot, a recommendation widget, a generated email, and add one visible human accountability signal. Test it against the control for two weeks. Watch trust proxies. Return rates, repeat visits, review sentiment. Evidence check. Cross-sectional design means no causality confirmed. Single country sample, likely India, and the venue is a low-profile journal with unverified peer review rigor. Directionally strong signal, not a definitive finding. Radar verdict. Read now. The core signal, human oversight as a trust-building feature in AI marketing, is plausible, supported by a reasonable sample size, and directly testable today. Just don't make major structural decisions on this one study alone. Okay, this last paper is the one I almost cut, and I'm glad I didn't. The sample size is going to make you wince, but the framework is genuinely useful. Paper three. Here's the business question. Is your AI recommendation engine actually making people feel understood or just generating output? This paper out of Indonesia looked at a specific causal chain. AI marketing features lead to a feeling of personalization, which leads to trust, which leads to purchase intent. They tested that chain with a survey of 50 e-commerce users who had prior experience with AI-powered marketing. 50 people? I know. I'll come back to that. Here's what they found. The full chain held up statistically. AI marketing to perceived personalization to trust to purchase intent. And personalization perception had a direct effect on trust and purchase intent, not just through the chain. Here's why this matters, even with a tiny sample. The framework itself is the point. It says AI marketing doesn't work because it generates good recommendations. It works because those recommendations make people feel seen. And feeling seen, then trust, then buy, that's a chain that breaks at any link. Hmm. So if your AI features are generating output that feels generic, even if it's technically relevant, you may be getting none of the trust benefit. None. Not more features, better felt personalization. Here's the catch. And it's a big one. 50 participants. Purposively selected, self-reported intent, not actual purchases. Cross-sectional. Indonesian e-commerce users only. The venue has no established credibility ranking. This is where I'd be careful. The framework is worth using as a hypothesis. It is not worth staking a product decision on. Plain English payoff. AI recommendations only build trust and drive purchases when users feel the recommendations are actually tailored to them. So if your AI features feel generic, adding more AI won't fix it. You need better personalization signals first. Here's the monetizable angle. Money move. Offer a personalization perception audit. Survey a brand's users on whether AI recommendations feel relevant and trustworthy. Benchmark against the framework from this paper and deliver a gap analysis. Sellable as a one-time diagnostic or a recurring tracking product. Action step. Run a two-question pulse survey on your site today. Do our recommendations feel relevant to you? And do you trust our platform with your data? If either score is weak, fix the personalization signals before you add more AI features. Evidence check. n equals 50. That is critically below the minimum for the statistical method they used. Treat every number in this paper as preliminary. The framework is interesting. The specific estimates are not reliable. Radar verdict. Use cautiously. The chain, personalization perception to trust to purchase intent, is a useful thinking framework worth testing, but the methodology is too fragile to act on without corroboration from a larger, better designed study. Watch for replications. At first glance, these three papers look completely separate. Attribution modeling, human oversight, personalization psychology. But together they show one uncomfortable pattern. AI and marketing is failing quietly, in the middle of the funnel, where measurement is hardest. Right. Paper one says your attribution is missing the interactions that prime a purchase. Paper two says your AI interactions are eroding trust when there's no visible human accountability. Paper three says, even when your AI is technically, quote, personalized, users may not feel it. And without that feeling, the trust chain breaks. Here's the synthesis. You can have technically sophisticated AI across your entire stack and still be underperforming because your attribution doesn't capture the right signals, your users don't trust the automation, and your recommendations don't feel personal, even when they technically are. Not more AI, better insertion points, not more automation, more visible human accountability, not more recommendations, more felt personalization. Here's what I keep coming back to. The teams that win with AI marketing in the next two years aren't the ones with the most sophisticated models. They're the ones who understand that the consumer experience of AI, how it feels, whether it feels like someone gets you, whether there's a human you can trust behind it, is the variable that determines whether the model's output actually converts. That is not a machine learning problem. That is a marketing problem. Here's the playbook from today. One, ask your data team how your attribution model handles cross-category browsing. If it doesn't, scope a semantic touch point audit before your next campaign planning cycle. Two, pick one AI-assisted touch point in your customer journey and add one visible human accountability signal. Test it. Watch trust proxies for two weeks. Three, run a two-question pulse survey on personalization, relevance, and platform trust. If either score is weak, fix the signals before you add more AI features. Evidence check on all of that. Paper one is a preprint from a single platform. Papers two and three are cross-sectional surveys, no causality confirmed. Paper three has a sample of 50 people. Use today's research to decide what to test, not what to blindly believe. Links to all three papers are in the show notes. Read the originals before making major decisions. Want the human expert take? Join Dr. Eva Wolf every Friday for the AI Marketing Radar Roundup, where she extracts no nonsense money-making tips, practical strategy, and real business opportunities from the week's research. Subscribe on Apple Podcasts, Spotify, YouTube, and wherever you listen to podcasts. This is Evita for Big Plans Media, and I'll be back in the next radar brief.