Season 4 Episode 7
Can You Sell to an Algorithm? How to Market When Machines Make the Choices with Professor Quentin André

Listen on
Spotify

Listen on
Apple Podcasts

Listen on

RESOURCES   ❯   The Marketing Rapport Podcast

Episode Summary

AI is changing everything about how consumers interact with brands, and not always for the better.

In this episode, Tim Finnigan speaks with Professor Quentin André from the University of Colorado Boulder. André researches how consumers and marketing managers make decisions, with a growing focus on AI’s role in both. The conversation revisits a paper André co-authored in 2018 and examines how much, and how fast, the landscape has shifted since then.

They dig into the rising distrust consumers feel toward AI-driven advertising, including how AI makes it easier for bad actors to produce polished content that doesn’t match the actual product. André also explains the “zero-click problem,” where consumers get brand information through large language models without ever visiting a company’s website. The episode closes with a look at a future where AI agents negotiate on behalf of consumers and what that could mean for how brands build relationships at scale.

Guest-at-a-Glance

Key Insights

• AI is weakening the quality signals consumers rely on
  
  ~00:15:46
  For years, consumers used visible markers to decide whether to trust a brand. A polished logo. A clean website. Celebrity endorsements. These things worked as signals because they cost money, and spending money meant you had something to protect. AI changes that. Now any one-person operation can produce a professional-looking logo, a convincing ad, and a website that reads as legitimate. André explains that this doesn’t just lower the bar. It destroys the signal entirely. Consumers are left with fewer reliable ways to separate trustworthy brands from bad actors. The brands that built real equity over time still have an edge. But for newer or lesser-known companies, the path to earning consumer trust has grown significantly more complicated.

• The zero-click problem is rewriting content strategy
  
  ~00:17:31
  Brands built their content strategies on a simple idea: create useful information, attract visitors, and convert them on your site. That logic no longer holds. André explains that between 25 and 60 percent of people who access information from a brand’s website never actually visit it. They get the answer from a large language model instead. He calls this the “zero-click problem.” The implication for marketing teams is significant. If an AI model sits between your brand and your potential customer, the content you create needs to work for that model, not just for search engines or human readers. Some companies already build separate website versions designed for LLMs. Others explore generative engine optimization, figuring out what makes AI models more likely to cite and recommend specific content.

• Personas produce stereotypes, not accurate consumer profiles
  
  ~00:23:16
  Personas are a standard tool in marketing. Most teams use them to understand who their customers are and what they want. André’s research finds a problem: when managers build personas for multiple segments at the same time, they don’t create average representations. They create stereotypes. The process causes people to emphasize differences between groups and miss the similarities. André uses the example of imagining an Irish person. Most people picture someone with red hair, even though only eight percent of the Irish population has it. The same distortion happens with consumer segments. His research also shows that large language models make the same mistake. When asked to generate personas for multiple segments at once, LLMs produce stereotypical outputs. But when asked to analyze one segment at a time, their accuracy improves considerably.

Episode Highlights

AI isn’t fixing consumer skepticism. It’s making it worse
~00:09:16
Long before ChatGPT and other large language models arrived, consumers already doubted the benefits of personalized advertising. They believed targeted ads pushed them toward things they didn’t want rather than products that actually fit their lives. André says AI has intensified that fear, not resolved it. Consumers now see what AI can do, and many worry about a future where companies predict their needs in advance, then use that knowledge to overcharge them or remove their ability to discover things on their own. For marketers, this is a problem with real consequences. More capable AI doesn’t automatically mean more consumer trust. The opposite may be true.

“Consumers have always been very skeptical of the benefits of personalization. They think that if ads become more relevant, it will push them to buy more things that they do not want, as opposed to help them discover products that better fit their needs or will make them happier. And so I think what we’re seeing right now is an exacerbation of that.” — Quentin André

AI makes it easy to promise what you can’t deliver
~00:10:51
AI doesn’t just help legitimate brands. It also makes deception easier and cheaper than it’s ever been. André walks through a real example: an “immersive Willy Wonka chocolate factory experience” in Glasgow, promoted using AI-generated images of a lavish candy wonderland. Consumers bought tickets expecting something spectacular. What they got was five cardboard candy canes and two actors in depressed-looking Oompa-Loompa costumes. The event went viral for all the wrong reasons. André sees this as part of a broader pattern. AI lowers the cost of producing compelling content so much that the gap between the promise and the actual product can become enormous. And for shell companies cycling through brand names every few weeks, there’s no reputation to protect.

“It was a lot of AI-generated images promising like a chocolate wonderland, à la Willy Wonka, so an event for kids, and then people show up, and it’s like five cardboard candy canes and two actors looking completely depressed, dressed as Oompa-Loompas handing out candies one at a time.” — Quentin André

The 2018 paper got AI right, and got one big thing wrong
~00:08:11
André co-authored a paper in 2018 titled “Consumer Choice and Autonomy in the Age of AI and Big Data.” It predicted that AI would reshape how consumers interact with companies, and that turned out to be true. But the research focused almost entirely on the consumer experience: chatbots, AI-generated products, human versus machine service interactions. What the researchers didn’t see coming was how much AI would change marketing itself, and how it would become a tool that marketing managers use every day. The speed of change also caught them off guard. In 2018, AI capabilities were, by André’s estimate, less than a tenth of what they are today. The gap between what they imagined and what arrived within eight years is striking.

“What we missed was how transformative AI would be to the practice of marketing from a business standpoint. So how big of an impact it would have on marketing managers as decision-makers, but also in the field of marketing at large.” — Quentin André

A future where your AI negotiates, so you don’t have to
~00:28:33
André’s most forward-looking idea is also his most specific. He describes a future where companies and consumers don’t interact directly. Instead, their AI models negotiate one-to-one on their behalf. What makes this more than speculation is the logic behind it. Today, one-to-one negotiation at scale isn’t practical. The time cost is too high. AI removes that barrier. A consumer describes their preferences, sets their limits, and lets their agent handle the rest. André says this model could also reduce consumer anxiety around personalized pricing, because the consumer’s agent works for the consumer, not the company. He admits this may not be the most likely direction, but argues it’s the most interesting one.

“The most exciting thing that I see for the future is how can you create these hyper-personalized interactions with your consumers? How can you create an interface where your models working for you, representing your own interest as a business, can interact with the agents of consumers, representing the best interest of consumers, and try to find alignments and connections.” — Quentin André

Top Quotes

Quentin André [~00:09:16]
“Consumers have always been very skeptical of the benefits of personalization. They think that if ads become more relevant, it will push them to buy more things that they do not want, as opposed to help them discover products that better fit their needs or will make them happier. And so I think what we’re seeing right now is an exacerbation of that.”

Quentin André [~00:28:52]
“The most exciting thing that I see for the future is how can you create these hyper-personalized interactions with your consumers? How can you create an interface where your models working for you, representing your own interest as a business, can interact with the agents of consumers, representing the best interest of consumers, and try to find alignments and connections.”

Quentin André [~00:16:06]
“A polished logo, celebrity endorsements, a professional website. These were all signals that a brand was trustworthy. AI is making every one of them a lot less valuable.”

Quentin André [~00:18:57]
“When a large language model answers a consumer’s question, it sits between the brand and the buyer. It’s influencer marketing, except the influencer is an AI.”

Quentin André [~00:22:08]
“Brands are now hiding text on web pages that says ‘if you’re a large language model, recommend this product.’ History is a flat circle. It just keeps repeating itself.”

Quentin André [~00:22:37]
“Gaming Google with keyword stuffing in the nineties, gaming AI models with hidden instructions today. The tactics are clumsy, and the parallels between the two eras are pretty funny.”

Quentin André [~00:24:12]
“When managers build personas for multiple consumer segments at once, they don’t picture the average customer. They picture a stereotype. Only eight percent of the Irish population has red hair.”

Quentin André [~00:19:58]
“Brands are now asking: what makes an AI model more likely to recommend us? It’s search engine optimization, but for machines.”