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 Customers, Chatbot Ads & Personalization Trust: 3 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 Evita, 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. What if the customer your marketing was built to reach isn't a human anymore? And what if the AI that is shopping for people, the one that actually picks the product and hits buy, doesn't care about your brand story, doesn't respond to emotional triggers, just parses structured data and moves on. That's not hypothetical. That's where today's papers are pointing, from completely different angles. We screened 388 papers. Three cleared the full text bar and made the radar today. Quick caveat: this is a first pass research briefing, not a final academic review. 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. If an AI agent, not a person, is doing the shopping, who exactly are you marketing to? SARSDET and colleagues published a conceptual piece in a marketing analytics journal this year. The argument is this AI systems aren't just tools that help humans shop anymore. They're becoming the actual shoppers. Think about it. Someone tells ChatGPT, find me a good running shoe under $150 and order it. The AI is the customer now, not the person. The person delegated. And your product listing, your brand story, your emotional hook. None of that was written for the thing that just made the decision. The authors lay out a spectrum. On one end, AI assistants that help humans decide but still need sign-off at every step. On the other end, fully autonomous AI agents that search, compare, and buy with little to no human input. The more autonomous the AI, the more it replaces the human as the actual decision maker. Right. Here's what that means in plain English. Your entire marketing stack, your copy, your targeting, your checkout flow was designed for human psychology, loss aversion, social proof, scarcity. None of that works on a machine evaluating structured product data. The authors call the solution machine marketing, a new subdiscipline focused on how to reach, influence, and communicate with AI systems buying on behalf of humans. Now, let me be clear about what this paper is and isn't. It's conceptual, no data, no experiment. It's a framework paper, asking a question more than answering one. The authors use real-world examples, AI planners, booking concierges, shopping assistants, but those are illustrations, not evidence. But here's the catch. Real-world adoption of fully autonomous AI agents completing purchases is still emerging. This paper may be a little ahead of market reality. That's not a reason to ignore it, that's a reason to get ahead of it. That's the part I keep coming back to. The brands that win in this next phase won't be the ones who figured it out after AI agents took over shopping. They'll be the ones who started optimizing for machine readers two years before it became obvious. Plain English payoff. Your product listings need to be as readable to an AI agent as they are compelling to a human. And right now, most of them aren't built for either job equally. Okay, here's where this becomes commercially interesting. Money Move. Build a machine readable product listing audit, an agency service or SaaS tool that scores and rewrites product pages for AI shopping agents. Think of it like SEO, but for ChatGPT's shopping layer. That service barely exists yet. That is where the money is. Action step. Go to ChatGPT or Gemini right now and ask for a product recommendation in your category. See if your brand comes up. If it doesn't, or if it does but the information is incomplete or wrong, you have your first audit priority. Evidence check. Conceptual paper. No data, no experiment, no user study. The machine marketing subdiscipline doesn't exist yet. They're arguing it needs to be created. Use this for strategic orientation, not tactical proof. Radar verdict. Read now. Strong framing, zero empirical evidence, but the question it asks is one every marketer needs to be sitting with today. This next one looks like a technical paper. And it is, but stay with me, the business implication is enormous and it's happening right now, whether your team is watching or not. Paper two, here's the business question. How do you run ads inside a chatbot response without making the answer worse and without rebuilding your AI from scratch? This is a preprint from a research team published in late July. They built something called Pila, plug and play insertion for LLM native advertising. And the core idea is elegant. Instead of baking ads into your main AI model, which is expensive and messy, you train a small secondary model to rewrite the AI's response after it's already generated. The main model does its thing. Then Pila steps in, takes that response, and weaves in the sponsored content. Clean, modular, no surgery on your primary model. They tested it against three existing approaches: prompt-based insertion, sampling-based, and fine-tuning the main model directly. PELA outperformed prompt-based methods by 34%, sampling-based by 47%, and fine-tuning by 8% on a combined score of ad quality and response naturalness. And here's the detail I think is genuinely interesting. Pila includes something they call an ad intensity controller. You can dial how prominently the ad appears, from subtly woven in, almost invisible, to more overtly promotional. That gives operators a real lever for the user experience versus advertiser exposure trade-off. That's not just a research feature, that's a product feature. When they put Pila on top of seven major commercial models, GPT, Claude, Gemini, and others, it improved the combined user satisfaction and ad effectiveness score by roughly 17 to 18%, without changing anything about how those models work underneath. But here's the catch, and it's a real one. No actual users were tested. The naturalness and ad effectiveness scores are benchmark metrics, not human reactions. We do not know if real people find these ads acceptable, intrusive, or deceptive. And the training data, 25,000 examples, was synthetically generated, not pulled from real advertising scenarios. And the disclosure question? Totally unaddressed. The paper doesn't touch regulatory rules, consumer consent, or what it means to insert sponsored content into an AI answer without flagging it. That is not a compliance footnote. That is a trust problem waiting to become a headline. What gets me is how close this already is to being real infrastructure. This isn't a theoretical sketch. They trained actual models, ran actual benchmarks across actual commercial LLMs. This is a blueprint. And someone is going to build it commercially. The question is whether they build the disclosure layer in from the start or bolt it on after the first regulatory complaint. Plain English payoff. There is now a working technical blueprint for inserting ads into AI chatbot responses without degrading answer quality. And the brands paying attention now will be the ones buying that inventory first. Okay, here's the business hiding inside the research. Money move. If you run an AI assistant or chatbot with real traffic, an ad insertion middleware layer, a PILA-style sidecar model, is a monetization path that generates ad revenue without rebuilding your core product. That is a new revenue line. Build it with disclosure baked in from day one. Action step. If you're buying placements in any AI-powered search or chat product right now, ask your vendor one question. Does your ad insertion degrade answer quality? And how do you measure that? If they don't have an answer, that's your signal to pressure test the placement before scaling spend. Evidence check, unreviewed preprint, synthetic training data, no real user study. The benchmark improvements are real, but they're technical metrics, not revenue, not user satisfaction, not advertiser ROI. Directional, not conclusive. Radar verdict. The concept is sound, the blueprint is clear, and the commercial urgency is high, but verify any claims with your own user research before committing budget. Okay, last one. I almost undersold this one when I first read it, so I want to land it right. Paper three. Here's the business question. You've invested in AI personalization. Is it actually making people buy more, or is it quietly eroding the trust that makes people want to buy from you at all? This is a qualitative systematic literature review. Suryati and Perulian 2026, synthesizing studies on AI-driven personalization, consumer trust, and purchase behavior across multiple major academic databases. Short version of what they found. AI personalization makes recommendations feel more relevant. But whether that relevance converts to purchases depends almost entirely on one thing. Whether the customer trusts you with their data. Not trust in a vague brand sentiment way. Trust in a specific functional way. Do customers feel their data is being handled fairly? Do they feel like they have some control? If yes, engagement goes up, purchases go up, loyalty goes up. If no, even the most precisely targeted recommendation pushes people away. And here's the mechanism. It's not a straight line from personalization to purchase. The chain is personalization, then trust, then engagement, then buying behavior. Skip trust, the whole thing breaks. Better engagement, actually feeling connected to a brand is what converts trust into revenue. Personalization alone does not get you there. Hmm, not more personalization, more trust. That's the finding. Now let me tell you what this paper is. It's a literature review. It synthesizes existing research. It cannot establish new causal relationships on its own. And the venue is a low-profile publication. The review methodology is underspecified. We don't know exactly how many studies were included or what the inclusion criteria were. So I'm not going to oversell it. But here's the thing, and this actually bothers me. The finding is confirmatory of something most marketers already know is true and almost nobody acts on. We keep adding personalization signals and we keep forgetting to add the transparency layer that makes people okay with it. I know. The review even identifies the practical fix. Show people why they're seeing what they're seeing. We recommend this because you browsed X. Give customers a visible way to adjust or opt out. Not buried in settings, visible. That single move, showing your work, can shift a customer from surveillance anxiety to feeling like you're paying attention in a helpful way. Plain English payoff. AI personalization only converts when customers feel their data is handled fairly and they have some control. And most brands are investing in the personalization while skipping the part that actually makes it work. Okay, here's where this becomes commercially interesting. Money Move. Offer a personalization transparency audit as a standalone agency service. Assess whether your client's AI targeting explains itself to customers, surface the trust gaps, and deliver a prioritized fix list. Concrete deliverable, easy upsell, and almost no brand has done this properly yet. Action step. Pull up your best performing personalized email or ad campaign and ask yourself, does it show the customer why they're seeing it? If the answer is no, add one line of context based on your recent visit, because you've been interested in X, run it against your current version before your next campaign review. Evidence check. Literature review with underspecified methodology and a low profile venue. No new data, no effect sizes, no causal claims. Useful for orientation. Don't use this as empirical proof for a budget decision. Radar verdict. Use cautiously. The insight is real, the evidence chain is solid in concept, but the methodology limits how far you can take it. Use it to sharpen your thinking, not to justify a platform rebuild. At first glance, these three papers look like they're covering completely different ground. One is about AI agents as shoppers, one is about ads inside chatbot responses, one is about personalization and trust. But together they show the same thing. The entire architecture of marketing is shifting from human to human to human to machine to human. And the gap between those two models is where the risk and the opportunity is sitting right now. Paper one tells you the customer may not be a person anymore. Paper two tells you there's already a technical blueprint for monetizing that new customer relationship. Paper three tells you that if you skip the trust layer with human customers or with the systems they're delegating to, the whole thing falls apart. Not more AI, better insertion points, not more personalization, more transparency, not a complete rebuild, a layer on top of what you already have. Here's what I keep coming back to. The brands that struggle in this next phase aren't the ones who ignored AI entirely. They're the ones who added AI to their existing human psychology playbook without asking whether that playbook even applies anymore. I'm telling you, that's the real gap. And here's the tension worth sitting with. Evidence check on all of that. One paper today is purely conceptual. One is an unreviewed preprint with no real user data. And one is a literature review with underspecified methodology. Use these papers 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 Wolfe 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.