AI & Marketing Research with Dr. Eva Wolf
Not another AI news podcast. This is a research radar — a twice-weekly briefing that surfaces peer-reviewed studies on AI and marketing, tells you what the evidence actually says, and helps you decide what's worth a deeper read.
AI & Marketing Research with Dr. Eva Wolf
AI Marketing Research: Consumer Trust, Bias & Chatbots
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Here's the uncomfortable question this week. If AI is writing your ads, optimizing your layouts, running your chatbots, how much of that is actually working the way you think it is?
SPEAKER_01That's the thread running through today's papers. Five studies, all poking at the same nerve.
SPEAKER_00We screened 120 papers. Five made the radar.
SPEAKER_01Quick caveat. This is a first pass research briefing, not a final academic review. We'll tell you what the papers suggest, what they don't prove, and which ones deserve a deeper read.
SPEAKER_00Okay, let's get into it.
SPEAKER_01Paper one. Here's the tension every marketer using AI content is about to run into. And most don't know it yet.
SPEAKER_00What's the setup?
SPEAKER_01You disclose that your ad was AI generated. Does that help you or hurt you?
SPEAKER_00And the answer can't just be it depends. That's not useful to anyone.
SPEAKER_01Right. So a team ran a systematic review, Prisma Protocol, the gold standard for this kind of synthesis. 35 studies on how consumers respond to AI-generated marketing content.
SPEAKER_0035 studies, that's not one survey of 200 college students, that's a map.
SPEAKER_01It is, and the map shows when consumers find out marketing content was made by AI, many of them trust it less. Two reasons. First, it feels less real. No human creativity, no genuine emotion behind it. Second, some consumers feel a low-level moral discomfort. Not quite disgust, but close.
SPEAKER_00Like a gut reaction. That's not a rational objection, that's visceral.
SPEAKER_01Exactly. And you can't logic your way out of a gut reaction.
SPEAKER_00Okay, but is this true for every type of ad?
SPEAKER_01No. And that's the piece that's actually actionable. Emotional ads take a much bigger hit from AI disclosure than rational, fact-based ads.
SPEAKER_00Say more.
SPEAKER_01If your ad is, here are the specs on this laptop, AI writing, that is fine. Nobody expects a spec sheet to have a human soul.
SPEAKER_00Right.
SPEAKER_01But if your ad is a heartfelt story about a mother and daughter and you slap AI generated on it, the whole thing falls apart. The review says exactly that.
SPEAKER_00And I want to flag something, because this isn't AI disclosure always hurts. The paper's clear, that audience matters too. Right.
SPEAKER_01Consumers who already know a lot about AI are less bothered. Cultural background matters. And if you make the AI feel more human, give it a name, a face. Some of that trust damage gets offset.
SPEAKER_00So the effect is real, but it's not a death sentence. It's a dial. Plain English payoff. Don't put an AI label on your heartfelt brand story. Save the disclosure for product specs and rational content where nobody expected human emotion in the first place. Money Move. Help brands A B test AI disclosure framing before they're forced to use it. AI assisted by our team versus fully AI generated probably lands very differently. That testing service doesn't really exist yet. Try this by Friday. Pull your three most emotionally driven campaigns. Ask if you had to label these AI generated, would they survive that? If the answer is no, that's where you need a human in the creative loop.
SPEAKER_01Evidence check. The underlying studies are almost entirely from 2024 and 2025. The field is brand new, and almost none of them measure actual purchases or clicks. They measure attitudes, how people say they feel.
SPEAKER_00So we don't know yet if the trust drop translates to a revenue drop.
SPEAKER_01Not yet. That gap is the next paper someone needs to write.
SPEAKER_00Radar verdict. Read now. It's the clearest practitioner map we have on AI disclosure risk. The emotional content warning alone is worth your time.
SPEAKER_01Paper 2. This one's a warning shot for anyone using AI tools to optimize how their ads look before they run them. Exactly those. Researchers compared AI attention predictions against actual eye tracking data from Brazilian consumers, real people, real visual stimuli in a lab. And the AI got it consistently wrong. The tool kept predicting people would look at the flashy, high contrast, visually bold elements. The stuff that pops.
SPEAKER_00Put the bright thing here.
SPEAKER_01Right. But the Brazilian consumers, they spent more time reading text, drawn to meaning, to context, to what the thing actually said, not what it looked like.
SPEAKER_00That's almost the opposite of what the AI predicted.
SPEAKER_01Almost exactly the opposite.
SPEAKER_00Okay. Here's why this matters beyond Brazil. These AI attention tools are trained on data, and that data is mostly American and European users.
SPEAKER_01So when you point the tool at an audience that processes visual information differently, the predictions fail.
SPEAKER_00You've built your creative around a map of the wrong territory. You've optimized for a ghost audience.
SPEAKER_01And the researchers make a point I think is underappreciated. When an ad is designed around wrong attention predictions, real consumers have to work harder to find what they need. Cognitive load goes up.
SPEAKER_00So not only is the ad not doing what you think, it might be actively making things harder for the person you're trying to reach.
SPEAKER_01That's the implication. Not directly measured, but a reasonable inference from the data.
SPEAKER_00Try this by Friday. Running campaigns outside the US or Western Europe. Pick one piece of creative your AI tool optimized. Run a five-person user test. Ask what they look at first. See if it matches the heat map.
SPEAKER_01Evidence check. One AI system, one market, Brazil. We don't know if the same gap shows up elsewhere, and sample sizes across the three eye tracking studies aren't fully reported. Use it as a reason to ask harder questions of your vendors, not as proof of universal failure.
SPEAKER_00Radar verdict, read now. If you're anywhere near emerging markets and you're using AI creative optimization tools, this paper should make you slightly nervous. Slightly nervous is the right reaction.
SPEAKER_01Paper 3, the most intuitive finding of the week. But there's a mechanism buried in it that changes how you should think about your AI marketing stack.
SPEAKER_00Hit me.
SPEAKER_01Survey study, 150 consumers with prior AI marketing experience. They looked at three AI qualities perceived effectiveness, trust, and whether the AI keeps learning over time. Then they measured engagement and satisfaction. And all three drive engagement, but here's the part that matters. AI doesn't directly produce customer satisfaction. It produces engagement first. And engagement is what produces satisfaction.
SPEAKER_00So the chain is good AI, engaged customer, satisfied customer, not good AI, satisfied customer.
SPEAKER_01Right. Satisfaction isn't a direct output of AI quality. It's downstream of the interaction.
SPEAKER_00That's actually meaningful for how you measure this. Because if you're asking, is our AI making customers happy? you're measuring the wrong thing.
SPEAKER_01You should be measuring whether it's making them interact. Click, respond, come back.
SPEAKER_00And the continuous learning piece that stood out to me too.
SPEAKER_01Yeah, consumers responded better to AI systems that seemed to improve over time. Tools that got smarter about their preferences.
SPEAKER_00Which makes sense. If a recommendation engine gives you the same slightly off suggestions every visit, it feels broken. If it seems to be figuring you out, you forgive a lot. Plain English payoff. AI doesn't make customers happy directly, it gets them engaged first. And engagement builds satisfaction. Optimize for interaction before you optimize for conversion. Try this by Friday. Look at your AI-powered recommendation or chatbot flow. What's your engagement metric, not your conversion metric? If you can't answer that, you're measuring the wrong thing.
SPEAKER_01Evidence check, 150 respondents. One country, convenience sample, self-reported perceptions. The venue is low credibility, directional, not definitive.
SPEAKER_00Use it to form a hypothesis, then test it on your own data.
SPEAKER_01Do not restructure your strategy around this one.
SPEAKER_00Radar verdict used cautiously. Okay, lightning round. Two more papers. Let's go.
SPEAKER_01Paper 4. AI-driven personalization and customer loyalty. The core claim: more personalized marketing leads to more loyalty isn't surprising. But the paper flags two things worth noting. Overpersonalization is a real risk. If customers feel surveilled rather than understood, trust drops and loyalty goes with it. And the paper argues human review points inside automated personalization systems are associated with better long-term outcomes.
SPEAKER_00Plain English payoff. Personalization builds loyalty until it feels creepy. The line between this brand gets me and this brand is watching me is closer than most AI tools acknowledge.
SPEAKER_01Evidence check. The methodology section is truncated in the available text. We can't verify sample size or analysis method. The venue is low credibility. Treat the findings as directional only.
SPEAKER_00Verdict.
SPEAKER_01More important than accuracy. More important than ease of use. Consumers want a fast answer before they want a right one.
SPEAKER_00So if your chatbot is slow and thoughtful, you've got the priority backwards. Fix the speed first.
SPEAKER_01Evidence check. A hundred respondents. Convenience sample. Single region in India. No inferential statistics. One data point, not a benchmark.
SPEAKER_00Verdict. Use cautiously. But the speed before accuracy finding is worth a quick test. If it holds in your market, it changes your chatbot optimization priorities completely. Okay, so here's what I think is actually happening this week. Every paper is a version of the same story. AI tools are being deployed at scale based on assumptions that haven't been validated for the specific audience context or culture they're aimed at.
SPEAKER_01The trust paper says disclosure effects depend on content type and audience. The attention paper says the optimization tools were calibrated on the wrong consumers. The engagement paper says the chain from AI to outcomes is longer than most people think.
SPEAKER_00Not one. They all say use it smarter, with local validation, with human review points, with an eye on the mechanism, not just the output.
SPEAKER_01And the evidence base is genuinely new. Most of these studies were published in 2024 and 2025. We're building the research plane while flying it.
SPEAKER_00Which means your competitive edge right now isn't we use AI. It's we actually check whether our AI is working. Here's the playbook from this week. One, audit your most emotionally driven AI content. If you'd have to label it A, I generated, does it survive that? Protect emotional campaigns with human creative leads. Two, ask your AI attention tool vendor what data it was trained on. If it doesn't include your actual market, run a local user test before you trust the heat map. 3. Shift one AI performance metric from satisfaction or conversion to engagement. Interaction first. See what changes. 4. Add a creepiness check to your personalization review. One person, one question. Does this feel helpful or does it feel surveilled? 5. If your chatbot's underperforming, check response speed before anything else. That's what consumers ranked highest.
SPEAKER_01Evidence check on all of that. Three of today's papers come from low credibility venues, and the evidence across all five is primarily self-reported attitudes rather than behavioral outcomes. Use this to decide what to test, not what to blindly execute.
SPEAKER_00Form the hypothesis, then run the experiment on your own audience.
SPEAKER_01Links to all five papers are in the show notes. Read the originals before making major decisions, especially the systematic review on AI disclosure. That's the strongest piece of evidence in this batch.
SPEAKER_00See you Thursday. And if something from this episode changed how you're thinking about your AI content strategy, I genuinely want to hear it. Find us.