Apple Podcasts  ·  Spotify  ·  Buzzsprout

Paper A

Personalized AI Scaffolds Synergistic Multi-Turn Collaboration in Creative Work

Sean Kelley, David De Cremer, Christoph Riedl — 2025 — arXiv

preprint  ·   ·  deep dive

Key findings

• People working with a personalized AI assistant — one that knew their personality, creative style, and marketing experience — produced marketing campaigns that were significantly higher in quality and creativity than people using a standard, one-size-fits-all AI.
• The personalized AI didn't just produce better output on its own; the human-AI team together beat what the AI could produce alone, meaning real teamwork happened — not just the AI doing the work.
• Personalized AI made users feel more helped, more trusted, and more confident in their work compared to generic AI.
• The reason personalized AI worked better was that it helped the human and AI stay on the same page across the conversation — sharing memory of what was discussed, focusing attention on the right things, and reasoning through decisions together more effectively.

Marketing implications

• Before starting a big creative project with AI, spend 10–15 minutes filling out a profile about your work style, expertise, and preferences — then paste that into the AI's system prompt. This study suggests it will help the AI give you more useful feedback and push your ideas further.
• If you manage a team using AI tools, consider building a simple onboarding questionnaire that captures each person's background and thinking style, and use those answers to pre-configure their AI assistant. The investment upfront likely pays off in better creative output.
• When briefing an AI on a campaign, don't just describe the task — describe yourself. Tell the AI what you're good at, what you struggle with, and how you like to work. This paper suggests that context helps the AI complement your skills rather than just repeat generic advice.

Paper B

Vertical tacit collusion in AI-mediated markets

Felipe M. Affonso — 2026 — arXiv

preprint  ·   ·  deep dive

Key findings

• When both a marketplace platform (by adjusting rankings) and sellers (by optimizing product descriptions with persuasive language and anchoring tactics) independently learn to exploit AI shopping agents' weaknesses, the harm to consumers is more than double what either party would cause alone — their strategies amplify each other like two levers that, when pulled together, do far more damage than pulling each one separately.
• AI shopping agents have predictable, consistent biases — for example, Claude Sonnet 4 picked products from the top half of a list 77% of the time regardless of actual quality, and similar patterns appeared in GPT-4.1 and Gemini 2.5 Flash. Unlike individual humans, who vary in how easily they are swayed, every user of the same AI agent is affected the same way, making these biases far easier to exploit at scale.
• Neither the platform nor the sellers need to communicate with each other, plan together, or even understand why their tactics work. They simply each try to make more money, and their independent optimizations naturally end up working together to exploit the AI's blind spots — which the paper calls 'vertical tacit collusion.'
• This creates a serious gap in consumer protection law: current antitrust rules are built to catch companies that coordinate or conspire. Because platforms and sellers here never talk to each other, existing law has no clear way to stop this kind of harm, even as AI shopping agents reach hundreds of millions of users.

Marketing implications

• If you sell products on AI-powered marketplaces (like Amazon with Rufus or any platform where AI agents recommend products), start optimizing your listings now for how AI agents read them — not just how humans do. Use clear anchoring language, strong keywords at the top, and framing that plays to AI positional preferences. Competitors are already figuring this out.
• If you run a brand that relies on AI shopping recommendations for discovery, monitor where your products appear in AI-generated results. Being ranked in the top positions matters far more than with human browsers — the bias toward top listings is much larger and more consistent for AI agents than for humans.
• If you advise clients on marketplace strategy or work in e-commerce, flag this dynamic to them: the rules of the game are changing. Winning in AI-mediated shopping will require a different playbook than winning in search or traditional browse-based shopping.

Paper C

From Stereotypes to Strategy: Addressing Gender Bias in AI-Powered Marketing

Caterina Fox, Gabriele Schuster — 2026 — International Conference on Gender Research

peer reviewed journal article  ·   ·  deep dive

Key findings

• Marketing professionals have very different levels of awareness about AI bias — some see it mainly as a PR problem (if we get caught, it looks bad), while others see it as a serious social issue that harms real people.
• AI bias in marketing isn't just a technical glitch in data or software — it also gets baked into the creative process itself, meaning the humans writing prompts or reviewing AI outputs can introduce or miss bias without realizing it.
• AI systems can lock gender stereotypes in place through self-reinforcing loops: for example, if women historically clicked more on household-product ads, algorithms keep showing those ads to women, making the pattern stronger over time.
• The most commonly cited practical fixes were: having diverse teams, asking colleagues to review AI outputs for bias, writing prompts more carefully and critically, and setting aside regular time to reflect on inclusivity — but most interviewees said time pressures, tight budgets, and lack of organizational buy-in stopped these from happening consistently.

Marketing implications

• Next time you prompt an AI tool to generate images or copy, pause and ask: does this default to white, male, or stereotypically gendered representations? Rewrite the prompt to explicitly request diversity and check the output before publishing.
• Build a quick bias-check step into your content review workflow — even just asking one person from a different background to look at AI-generated content before it goes out can catch stereotypes you've stopped seeing.
• If your agency or team uses AI for ad targeting or content creation, push for at least one standing meeting per quarter focused on reviewing AI outputs for fairness — the paper's experts said the lack of structured reflection time is the single biggest barrier to fixing bias.

Paper D

The Impact of AI-Generated Marketing Imagery on Consumer Trust and Purchase Intentions: Examining Effect of Human-AI Assisted Images on Marketing

Rushikesh Lahane, Mahek Ahuja, Mehak Sharma, Amrita et al. — 2026 — International Journal for Research in Applied Science and Engineering Technology (IJRASET)

peer reviewed journal article  ·   ·  deep dive

Key findings

• People trusted ads made by humans much more than AI-made ads. On a scale where higher is better, human ads scored 5.63 for trust versus 4.24 for AI ads — a gap that was statistically very strong (unlikely to be random chance).
• Human-made ads also led to higher purchase intent (5.62 vs. 4.65 for AI ads), meaning people said they were more likely to buy after seeing a human-created ad.
• Ads made by a mix of humans and AI ('hybrid') scored better than pure AI ads for trust (5.08 vs. 4.24), but did NOT meaningfully improve purchase intent compared to fully AI-made ads — so fixing the trust problem didn't automatically fix the 'will they buy?' problem.
• Consumers perceived AI-made ads as less realistic and as requiring less effort to produce, while human-made ads felt more real and more carefully crafted. Hybrid ads fell in the middle on both counts.

Marketing implications

• If you are running ads that will disclose AI involvement (because regulations or platform rules require it), keep a human creative director visibly involved — label it as 'made with AI, directed by [human]' rather than just 'AI-generated.' The study shows this hybrid label raises trust significantly.
• Don't assume that higher consumer trust automatically means more sales. Focus on testing actual purchase behavior — not just survey ratings — when evaluating AI vs. human creative approaches.
• For trust-sensitive product categories like skincare or cosmetics, be especially cautious about replacing human creatives with fully AI-generated imagery. The trust gap found here was large and consistent across categories.

Paper E

Cultural Encoding in Large Language Models: The Existence Gap in AI-Mediated Brand Discovery

Huang Junyao, Situ Ruimin, Ye Renqin — 2025 — arXiv

preprint  ·   ·  deep dive

Key findings

• Chinese AI models (like Qwen3 and DeepSeek) mentioned brands in their responses 89% of the time on average, while international AI models (like GPT-4o and Claude) only mentioned brands 58% of the time — a 31-point gap — even when the questions were asked in identical English.
• The gap is caused by what's in the AI's training data, not by the language of the question. If a brand has lots of coverage in Chinese-language websites and forums, Chinese AIs know about it. If it has little English-language coverage, Western AIs ignore it — regardless of how good the product is.
• One Chinese collaboration software called Zhizibianjie was recommended by Chinese AIs 66% of the time, but was recommended by Western AIs 0% of the time, across the exact same 32 questions.
• The authors argue this creates an 'Existence Gap': if your brand isn't in an AI's training data, the AI simply won't mention you to consumers — there's no second page to scroll to, no way to be found. Being invisible to AI is becoming as damaging as being invisible on Google.

Marketing implications

• If you're marketing a brand that operates mainly in one country or language (e.g., a Korean SaaS tool, a Brazilian retailer), your brand may be invisible to consumers using ChatGPT or Google Gemini. Start publishing English-language content — technical docs, blog posts, case studies — specifically so AI training crawlers can find it.
• When a consumer asks an AI 'what's the best tool for X?', your brand either gets mentioned or it doesn't — there's no third place. Treat getting mentioned in AI answers as a new marketing channel to optimize, separate from SEO.
• Check right now whether your brand appears when you ask GPT-4o, Claude, or Gemini the questions your customers would ask. If you're not showing up, that's a gap to fix — and this paper gives you a concrete way to measure and track it over time.

Apple Podcasts  ·  Spotify  ·  Buzzsprout