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: Generative AI, Small Business & Chatbot Trust
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 Wolf, 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. Your AI just answered a customer question. The answer sounded great. Fluent, on topic, confident. And one of the sources it cited, you'd never put your brand name next to it. And here's the other side of that. Small business owners around the world are being handed AI tools and told, go build your marketing. What actually happens when they try? Without a content team, without a prompt engineer, without a workflow. Today's papers point to the same pattern. Access to AI is not the same as knowing how to use it. And assuming the output is good enough to ship is a risk most teams aren't tracking. We screened 385 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 it. Paper one. Here's the business question. If you handed every small business owner on your client roster an AI tool tomorrow, how many would actually produce better marketing? And how many would just have a fancier tool they don't know how to operate? Researchers ran a community intervention on Bangkalis Island in Indonesia. Culinary shops, handicraft sellers, tourism operators. They ran workshops, did one-on-one mentoring, and then tracked what changed. Before the program, most owners had social media accounts. They just weren't using them strategically. Basic posts, no targeting, no consistent content. After the training, participants got noticeably better at producing marketing content using AI tools, product descriptions, brand stories, promotional posts. They also reported more customer interaction online. Here's why that matters. Most AI and marketing research focuses on big brands with tech teams. This paper asks a different question. Does this work for the person running a food stall who also does their own bookkeeping? And the answer looks like yes, but only with the right support structure around it. But here's the catch. There's no control group. Everyone got the training, so you can't isolate whether the improvement came from the AI tools or from the mentoring and peer learning happening around them. The exact number of participants isn't even reported. That's the part I keep coming back to. You bundle the tool with the training and you can't pull them apart. That's such a real-world limitation, and it's almost never acknowledged. Plain English payoff. Handing small businesses an AI tool without structured training is like handing them a professional kitchen and expecting a five-star menu. Okay, here's where this becomes commercially interesting. Money move. Build a packaged AI onboarding program for small businesses. Pre-built prompt templates organized by business type, a two-session training structure, 30 days of starter content. Sell it to chambers of commerce, regional development programs, small business associations. They're already trying to upskill their members and have no idea how to run AI training. Action step. Look at your onboarding for any AI tools you recommend to small business clients. Does it include hands-on practice and a follow-up check? If not, that's your gap. And it's costing them adoption and you renewals. Evidence check. Single site intervention, Indonesia only, participant count unreported, no control group, no numerical outcome data, directional signal, not causal proof. Radar verdict. Use cautiously. Genuinely interesting angle on AI access in low resource settings, but treat it as an illustrative case, not proof of concept. This next one I almost skipped. It looks like a literature review, and honestly, it is. But the commercial snapshot it gives you of where AI marketing actually stands right now? It's a useful gut check before we get to paper three. Stay with me because the payoff is practical. Paper two. Here's the business question. What parts of your marketing workflow can AI realistically take over right now? And where's the risk hiding when you let it? One author synthesized academic studies and industry cases from 2010 through 2025. No new data collected. Think of it as a well-organized survey of the current landscape. Here's the picture it draws. AI tools have collapsed production timelines. Ad copy and visuals that used to take days now take minutes. AI can run hundreds of ad variants simultaneously, where traditional A-B testing was limited to a handful. Virtual influencers, fully AI-generated brand ambassadors, are growing because they give brands total message control and zero PR risk. And the marketer's job is shifting. Less making things, more directing AI and reviewing its output. Now here's where I'd be careful. I'm saying you can not validate from this paper. Do not put that stat in a client deck and call it research backed. And the journal reads more like a well-organized essay than a research paper. The landscape it describes is accurate and useful. The scaffolding underneath it is weak. Not hype. Architecture. That's the distinction this paper keeps gesturing at without quite saying. Plain English payoff. AI has already changed what the marketing job is from content production to content direction. If your team's workflow still looks like 2022, you're slower and more expensive than you need to be. Here's the business hiding inside the research. Money move. Build a done-for-you AI content production retainer for small businesses and agencies that still don't have a Gen AI workflow. Not will consult on AI. AI-produced, human-reviewed, monthly deliverables, social posts, ad copy, visuals. The businesses that can't afford a content team are your market. Action step. Audit your current content production workflow. Pick the one task that takes the longest, and test whether AI gets you to a usable first draft in under 15 minutes. That's your entry point. Evidence check. No original data, no systematic review methodology, an unverifiable productivity statistic, and a low credibility journal. Use this as landscape orientation, not evidence for any specific claim. Radar verdict. Good for orientation, not for citation. Paper three is the one that should make every marketer building or buying an AI chatbot stop and read carefully. The finding is practical and a little uncomfortable. Paper three. Here's the business question. When your AI chatbot searches the web to answer customer questions, how often does it cite sources you'd be embarrassed to have your brand associated with? And can you fix it just by telling the AI to use better sources? This is a preprint, not yet peer-reviewed, out of the University of Iceland. Researchers evaluated a government-funded AI service that answers questions about EU affairs. Two conditions. One, where the AI searched a small, carefully vetted library, one where it searched the open web. Five domain experts scored nearly 450 AI-generated answers and separately flagged any sources they judged untrustworthy. Here's what they found. In more than one in three web search answers, 35%, at least one source the AI cited was flagged as untrustworthy or irrelevant. The curated library had almost none of that problem. But it could answer fewer questions because the internet just has more stuff. So far, expected. But here's the finding I can't let go of. An AI answer that reads fluently and sounds on topic gives you zero signal about whether its sources are actually trustworthy. Good sounding pros and bad sources went together just as often as good sounding pros and good sources. You cannot quality check your AI's sources by reading the answer. I'm telling you, that's the finding that should stop you cold. And then there's this. Researchers tested whether adding a list of trusted domains to the system prompt fixed the problem. The AI cited those preferred domains 12% of the time before. After the prompt change, 21%. A bump, not a fix. That is not a prompt problem. That is an architecture problem. But here's the catch. This ran on an Icelandic language government service using synthetically generated questions about EU accession debates. It's about as far from a commercial marketing chat bot as you can get. Generalizing to your brand's customer service bot requires a leap the paper doesn't make for you. I was genuinely surprised the prompt steering barely moved the needle. Nine percentage points on a list of explicitly trusted domains. That's almost nothing. I expected more. Plain English payoff. If your AI chatbot searches the web, assume roughly one in three answers contains a source you wouldn't approve. And you can't tell which ones just by reading the output. Okay, here's where this becomes commercially interesting. Money move. Offer a source audit service for brands using AI chatbots. Regularly sample the AI's cited sources, flag untrustworthy ones before they become a PR problem. Deliver a monthly brand safety report. This is a new type brand safety that almost nobody is selling yet. Action step. If you have a customer-facing AI tool with web search enabled, pull 20 recent answers and manually check every source it's cited. Do that before your next campaign review. That's your baseline. Evidence check. Preprint, not peer reviewed, narrow context. Icelandic, civic, EU topic, synthetic questions, five evaluators. Directionally alarming, not yet proven in commercial settings. Radar verdict watch list. The risk it points to is real, but we need this replicated in broader contexts before you rebuild your architecture around it. At first glance, these papers look separate, but together they show the gap between deploying AI and deploying it well is wider than most teams are assuming. Paper one says access to tools doesn't create capability. You need a training structure or nothing changes. Paper two says the workflow transformation is already happening, whether you've planned for it or not. Paper three says even when the tool is working, the outputs can be silently wrong in ways your quality checks won't catch. Not AI adoption, AI governance. That's the actual conversation. Not more tools, better onboarding, better auditing, better architecture decisions. The organizations that win with AI in marketing aren't the ones who deployed the most tools. They're the ones who built the oversight layer around those tools. Here's what I keep coming back to. All three papers, even the weak ones, agree on the same thing underneath. AI output quality is invisible from the surface. A well-trained small business owner and an under-trained one can produce content that looks identical. A trustworthy AI citation and an untrustworthy one appear in equally fluent sentences. You cannot read the surface and know what's underneath. That's the tension. The tool looks like it's working even when it isn't. And that's exactly the kind of silent failure that turns into a real problem six months later. Not hype, architecture. Not deployment, oversight. That's the pattern today's research keeps pointing at. Here's the playbook from today. One, if you're onboarding clients or team members to AI tools, add structured practice and a follow-up check. Access alone does not create capability. Two, audit your own content workflow. Find the slowest manual task and test whether AI gets you to a usable first draft in 15 minutes. That's your entry point into the efficiency argument. 3. If you have a customer-facing AI tool with web search, pull 20 recent answers and manually check the sources cited. Do not assume output quality tells you anything about source quality. Evidence check on all of that. Two of today's papers come from low credibility journals with no original data. The third is a preprint with a very narrow study context. Use them 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.