Apple Podcasts  ·  Spotify  ·  Buzzsprout

Paper A

Towards Gaze-Informed AI Disclosure Interfaces: Eye-Tracking Attentional and Cognitive Load While Reading AI-Assisted News

Pooja Prajod, Hannes Cools, Thomas Röggla, Pablo Cesar et al. — 2026 — arXiv

preprint  ·   ·  test this week

Key findings

• Short, one-line AI disclosures (like 'This article was AI-assisted') actually made readers work harder visually — they spent longer staring at parts of the text and scanned back and forth more — compared to readers who saw no disclosure or a detailed one. This was especially true when AI had only edited the article (rather than written parts of it).
• Detailed AI disclosures did NOT make reading harder — despite containing more words and information, they did not increase visual effort compared to having no disclosure at all.
• Neither the self-reported workload scores (NASA-TLX) nor pupil size changed meaningfully across any condition, meaning readers did not feel mentally taxed by any type of disclosure. The extra visual effort from one-line labels is something readers do automatically without noticing it.
• In interviews, readers strongly preferred detailed disclosures or 'show me more if I want it' designs over short labels — they wanted enough context to understand AI's role, not just a vague flag.

Marketing implications

• If you publish AI-assisted content (ads, blog posts, sponsored articles), don't use a vague one-liner like 'AI-assisted.' It actually makes readers scrutinize your content more intensely without satisfying their curiosity. Either say nothing or explain exactly how AI was used — both produce less visual friction than the vague middle ground.
• If your content platform or media buy includes AI transparency labels, push for 'expandable' or 'detail-on-demand' disclosure formats. Readers prefer them, and they don't create extra reading friction the way short labels do.
• If you're designing landing pages, sponsored content, or branded journalism that uses AI, treat the disclosure as content — write it properly rather than slapping on a badge. A sentence explaining 'AI helped structure this article; all facts were verified by our editors' is better than just 'AI-assisted.'

Paper B

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

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

preprint  ·   ·  use cautiously

Key findings

• Chinese AI systems (like Qwen3, DeepSeek, Doubao) mentioned brands 30.6 percentage points more often than Western AI systems (like GPT-4o, Claude, Gemini) — 88.9% vs. 58.3% — even when the question was asked in English. This means the AI's training data origin, not the language you use, determines which brands it suggests.
• A brand called Zhizibianjie (the authors' own company) was recommended 65.6% of the time by Chinese AIs but 0% of the time by Western AIs when asked the same questions in English. Good products can be completely invisible to AI if there's not enough written about them in the language that AI was trained on.
• The researchers call this the 'Existence Gap': if a brand isn't well-represented in the text an AI was trained on, the AI simply won't mention it — no matter how good the product is. There's no equivalent of 'page 2' in AI recommendations; if you're not mentioned, you don't exist.
• The researchers propose that brands should treat English-language content (technical docs, case studies, forum discussions) as a strategic asset that gets them 'encoded' into Western AI systems, similar to how SEO once got brands ranked on Google.

Marketing implications

• If you sell software or services and want Western AI chatbots (ChatGPT, Claude, Gemini) to recommend your brand, focus on publishing English-language technical content: detailed documentation, case studies, forum answers, comparison articles, and press coverage. Quantity and quality of English-language text about your product appears to be what gets you into AI training data.
• If you're a non-Western brand trying to enter Western markets, treat 'AI visibility' as a distribution problem, not just an SEO or advertising problem. Start building an English-language content presence now, because AI systems are trained on historical data — the content you publish today shapes whether future AI versions recommend you.
• If you run an agency, you could audit a client's AI mention rate across major LLMs (just run test queries) and sell 'GEO gap analysis' as a service alongside traditional SEO audits.

Paper C

A Study on AI-Driven Marketing and its Impact on Consumer Purchasing Behavior

Lokeshwari S, P. Brindha — 2026 — REST Journal on Banking, Accounting and Business

peer reviewed journal article  ·   ·  watchlist

Key findings

• AI-powered personalized product recommendations tend to increase customer satisfaction and push people toward buying — customers are more likely to purchase when shown suggestions tailored to them.
• The paper claims AI models can predict what a consumer will buy next with about 90% accuracy — though this figure is stated without a specific source citation in the text.
• Chatbots and virtual assistants, available around the clock, have improved how customers feel about brands by handling service requests quickly and conversationally.
• Younger people are more comfortable engaging with AI-driven marketing than older consumers, suggesting age matters when deciding how aggressively to use AI tools.

Marketing implications

• If your brand uses a chatbot, make sure it is available 24/7 and sounds human enough to resolve issues — this paper reinforces that responsive AI customer service improves how people feel about your brand.
• If you are targeting younger audiences (Gen Z, younger Millennials), lean into AI-personalized experiences like dynamic recommendations or tailored offers — they are more open to it than older customers.
• Whatever AI tools you use for personalization, publish a plain-language data policy on your site and be upfront about how you use customer data — the paper highlights that hidden or opaque AI practices erode trust and can backfire.

Paper D

Integrated AI-Driven Marketing Growth Models for Scaling Businesses in Competitive Direct-to-Consumer Landscapes

Dineth Ratnayake — 2026 — Zenodo (CERN European Organization for Nuclear Research)

peer reviewed journal article  ·   ·  use cautiously

Key findings

• Companies that deeply embedded AI across their marketing operations — not just in one-off tools — grew revenue faster and held up better against competitors than companies using AI in isolated ways.
• AI neural networks were better than traditional statistical models at predicting how much a customer would spend over their lifetime, which is a key number for planning marketing budgets.
• When firms were grouped by how advanced their AI use was, the most advanced group consistently outperformed the others on growth and resilience metrics.
• AI's effect on revenue growth worked partly through two steps: better customer data understanding first, then more accurate personalization — meaning AI doesn't just help directly, it also makes other marketing systems smarter.

Marketing implications

• If you run a D2C brand, don't just bolt on one AI tool (like a chatbot or an ad optimizer) and call it done — this study suggests the payoff comes from connecting AI across your customer data, personalization, and campaign systems together.
• If you're choosing between a traditional forecasting spreadsheet and an AI-based customer lifetime value model, this paper supports trying the AI model — neural networks outperformed conventional approaches for that specific prediction task.
• Before investing in more AI tools, audit which parts of your stack are already connected: are your customer data insights actually feeding your personalization engine? If not, that's the gap to close first.

Apple Podcasts  ·  Spotify  ·  Buzzsprout