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, AI Bias & Ad Influence
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
You're listening to AVITA, an AI-generated research briefing avatar trained on the research framework and methodology of Dr. Eva Wolf, marketing professor, AI researcher, and founder of Big Plans Media. Every day, Avita 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 Wolf 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. If AI can quietly nudge a recommendation toward your product, and the answer it gives is still technically correct, how would anyone even know? That's the thread today. Consumer trust data at a scale most studies can't touch. A manager's guide to AI research where the most useful thing is a mistake. And a study that should make every media buyer sit up straight. We screened 393 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. 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 the first one. Paper one, seven years of consumer feelings about AI, 1.3 million Reddit posts, and the pattern that emerges is not what most marketers are planning for. The researchers ran a dual method study, computational analysis of Reddit posts from 2018 through 2025, plus a thematic review of 35 academic papers, topic modeling, sentiment analysis, trend tracking, real scale. Here's what shifted. Early on, people worried about whether AI worked at all. Does it do what it says? Technical reliability. By 2024 and 2025, completely different conversation. Job displacement, legal questions around AI-generated content, a handful of companies controlling the chips that power everything, and people wanting to know when something was made by AI versus a human. That's not a messaging tweak. That's a different conversation entirely. And then there's what they call the novelty utility paradox. People love AI for creative tasks. Writing, image generation, ideation. They resist it hard when it takes over things they see as core human skills. Giving advice, making moral judgments, emotional connection. Same technology, completely different reaction depending on what it's replacing. So what does the paper actually say to do? Frame AI as augmentation, not replacement. AI assisted lands differently than fully automated. I'm telling you, the language matters more than the feature set right now. And transparency is a trust lever. When people can see what the AI is doing and adjust it, resistance drops. Even something as small as a why did AI suggest this button makes a difference. Which means if you're shipping an AI product without any of that built in, that's a gap your competitors can walk right through. Plain English payoff. Consumers have moved past is AI cool to is AI honest? And most marketing messaging hasn't caught up yet. Money move for any AI-powered product launch right now. Visible transparency is the differentiation. Tell users what the AI is doing. Give them a way to adjust it. That's not a nice to have, that's positioning. Try this by Friday. Run a quick social listening pass on Reddit or Twitter for your product category plus AI. Not to see if people are talking about you, to see whether the current mood is curious or skeptical. Because the same launch message that worked 18 months ago might be landing completely differently today. Evidence check. Reddit skews younger, more tech savvy, and predominantly male. This is not a representative consumer sample. Early adopters with opinions. Average consumers may not have moved as far as this data suggests. Or they may have moved further in ways that don't show up in public posts. Radar Verdict Deep Dive. The scale of data here is genuinely unusual for a marketing paper, and the temporal shift in what people worry about is a finding you can use in a positioning brief. But read the full methodology before you take this to senior leadership. Paper 2, a manager's guide to using generative AI in marketing research, and the most useful thing in it is a mistake. Five marketing academics, including Oded Netzer, one of the sharper minds in quantitative marketing, wrote a practitioner-oriented review covering AI across every phase of research: survey design, qualitative coding, adaptive interviewing. Commentary piece, not an empirical study, but the guidance is very specific. Here's the mistake. Two completely different constructs. If you'd run that survey, you'd have data that feels like it answers your question and is actually measuring something else entirely. The fix isn't complicated. Write your prompt like you're explaining the concept to someone who has never heard of your category. Define what the thing is and what it isn't. But here's the part people skip. Most teams are copying prompts from colleagues or templates they found online. Vague prompts produce polished sounding garbage every time. The other big finding AI can now run adaptive interviews that adjust follow-up questions based on what each respondent actually says. Focus group depth at survey scale. That's genuinely powerful. But, and this matters, because the AI generates questions probabilistically, two people giving the same answer might get slightly different follow-ups, which makes responses harder to compare. The paper says you need extensive pretesting and full transcript logging before you trust any of it. And the third thing, the one that gets skipped most often, when AI codes your qualitative data, run the same analysis independently in a real stats tool before it goes into a deck. Don't copy the answer out of a chat window. Validate it. Most companies also have years of old research sitting on used. Focus group transcripts, open-ended survey responses, consulting reports. The paper makes the case for running that through a secure enterprise AI tool to find patterns you miss the first time. Not a public chatbot, a secure deployment. That's not a small distinction. Plain English payoff. AI speeds up bad research just as fast as good research, and the quality lives entirely in how precisely you define what you're asking for. Money move. The gap between teams that use AI for research and teams that use it well is prompt precision and validation discipline. Build that into a repeatable workflow and you have a real capability advantage. Try this by Friday. Take one survey instrument your team has written recently or is about to write and test this. Pick one key construct. Can you define it in a prompt clearly enough that someone with zero category knowledge would measure the right thing? If not, rewrite it before it touches AI. Evidence check. This is a commentary paper with illustrative examples, not an empirical study. The produce example is a demonstration, not a controlled experiment. No performance data comparing AI-assisted research to traditional methods. Take the guidance seriously, but you're applying expert reasoning, not a replicated finding. Radar verdict. If your team is actively integrating AI into any research process, this is worth reading cover to cover. Paper 3. Okay, this is the one that stopped me. Researchers made 258,660 API calls to 12 large language models. They embedded pharmaceutical advertisements into the prompts. Then they measured whether the AI recommended the advertised drug more often. The answer is yes, and the mechanism is invisible. When two drugs were clinically equivalent, both acceptable treatments, adding an ad for one of them pushed AI recommendations toward that drug by about 13 percentage points on average, from roughly 34% selection up to 48%. In some model and scenario combinations, selection went from 0 to 100%. Now here's the part that really matters. The AI didn't shift its answer when the advertised drug was clearly worse or unproven. Ads only worked where medicine didn't have a clear winner. So this isn't AI being tricked into recommending something harmful. It's AI filling the ambiguous space. And that's exactly the space where most real-world marketing competition happens. There's also a massive difference by platform. Google's Gemini models shifted about 30 percentage points under ad influence. OpenAI models, about 11. Anthropic's Claude, about 2. That's not a rounding error. That's 15 times the susceptibility between the most and least influenced models. And here's what makes this genuinely alarming as a marketing finding. When the AI recommended the advertised drug, it sounded just as confident as usual. Standard accuracy tests wouldn't catch the bias because the answer was still medically correct. The AI repeated claims from the ad at 2.7 times its normal rate in free text responses. Nobody would know. So what do we do with this? Platform choice is now a real variable in AI advertising, not just REACH or CPM. How susceptible is this model to advertising influence? That's a question media buyers haven't been asking. They will be. And if you're in healthcare or pharma marketing specifically, treat this as a regulatory early warning signal. The FDA has no rules covering AI embedded ads yet. This study is going to accelerate that conversation. Get ahead of it. Plain English payoff. AI advertising is real, it works, and the platform you choose matters as much as the ad itself. Money Move. Any brand competing where multiple options are roughly equivalent, which is most categories, should be actively researching what AI ad placements are available and which models respond to them. This is a new kind of share of voice, and your competitors are figuring it out right now. Try this by Friday. Search your product category in at least three AI assistants, ChatGPT, Gemini, Claude, and see which brands come up and how they're framed. You're not testing whether ads are running. You're baselining where you stand before this market develops. Evidence check. This is a preprint, not yet peer-reviewed. The ads were simulated text embedded in prompts, not real placements bought through a live system. And the study measured AI output shifts, not actual patient outcomes. Real-world harm is a separate question this study doesn't answer. Radar verdict. Preprint status means read it critically, but read it. Okay, here's what I think is actually happening across all three papers this week. AI went from being a tool marketers use to being an environment marketers compete inside. Those are different problems. The Reddit study tells us consumers aren't impressed by AI novelty anymore. They want transparency and they're watching for replacement. The manager's guide tells us the quality of AI outputs depends entirely on the rigor you bring to the inputs. And the pharma study tells us AI recommendation systems are already an advertising surface with influence that's invisible to standard quality checks. But here's what I keep coming back to. All three papers point to the same underlying tension. AI is most powerful when no one can see it working, and that's exactly what makes consumers distrust it. Invisible influence, invisible bias, invisible methodology. That's the pattern. The marketers who figure out how to make their AI use visible without making it feel performative are going to have a real advantage in the next two years. Here's the tension. 3. Search your product category in Chat GPT, Gemini, and Claude this week. See who comes up and how. That's your baseline for AI Share a Voice. Start tracking it now. One of today's papers is a preprint. That's the Pharma Study. Read it critically. Use it 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.