Apple Podcasts  ·  Spotify  ·  Buzzsprout

Paper A

GEO: Generative Engine Optimization

Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan et al. — 2024 — ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2024)

conference paper  ·   ·  deep dive

Key findings

• Adding statistics, data points, and direct quotes from credible sources to your web content made those pages get cited up to 40% more often by AI search engines — when those pages were already in the engine's pool of candidate sources.
• What works depends on the topic: for factual questions (science, data-heavy topics), adding numbers and quotes helped most. For opinion or advice topics, using confident, authoritative-sounding language worked better.
• The same optimization techniques also worked on a real AI search engine (Perplexity.ai), boosting citation visibility by up to 37%, not just in the lab setting.
• Traditional SEO thinking (ranking higher in a link list) does not translate directly to AI search engines — visibility in AI answers is multi-dimensional, involving how much of your content gets quoted, where it appears, and how relevant it seems to the specific question.

Marketing implications

• When writing content for your website or blog, add real statistics, cite credible external sources with links, and include direct quotes from experts. These are the specific changes most likely to get your content cited by AI search tools like Perplexity or Bing Chat.
• Match your content style to the type of question you want to rank for: if you're targeting how-to or factual queries, load the page with data points and numbers. If you're targeting opinion or recommendation queries, write with clear, confident authority.
• Treat AI search optimization as a separate layer on top of traditional SEO — your page still needs to get crawled and retrieved first (traditional SEO handles that), but once it's in play, the writing style and content choices in this paper can affect whether an AI engine quotes you or ignores you.

Paper B

Machine marketing: rethinking the customer in the age of generative AI

Marko Sarstedt, Susanne Adler, Monika Imschloss — 2026 — Journal of Marketing Analytics

peer reviewed journal article  ·   ·  read now

Key findings

• AI tools like ChatGPT are no longer just helpers that answer questions — they are starting to act like customers themselves, searching for products, comparing options, and even completing purchases on behalf of users without the user needing to be involved at each step.
• There is a spectrum from 'AI assistants' (which help you decide but need your input along the way) to 'AI agents' (which handle entire shopping tasks on their own). As AI moves toward the agent end of this spectrum, it effectively becomes the buyer — not just an advisor.
• AI systems do not think like humans. Research in 'machine psychology' shows they follow different decision patterns, meaning marketers cannot simply assume that what works on a person will work on an AI buyer. They need to learn how to market to the AI itself.
• The authors argue that a whole new branch of marketing — which they call 'machine marketing' — needs to be created to study, predict, and influence how these AI buyers behave, just as traditional marketing studies human consumers.

Marketing implications

• If you sell products online, start thinking about whether your product pages and pricing are legible to an AI agent doing comparison shopping on someone's behalf — not just readable by a human. Structured data, clear specs, and consistent pricing matter more than ever.
• Test whether your brand shows up favorably when someone asks ChatGPT or a similar AI assistant to recommend products in your category. This is the new version of 'checking your Google ranking' — some people are already calling it Generative Engine Optimization (GEO).
• If you run consumer research or surveys, consider that AI-assisted respondents may give different answers than unassisted ones — factor this into how you design and interpret research going forward.

Paper C

PILA: Plug-and-Play Insertion for LLM-native Advertising

Zhaowei Zhang, Yuhan Fu, Yihang Zhang, Xiaohan Liu et al. — 2026 — arXiv (Cornell University)

preprint  ·   ·  read now

Key findings

• PILA can insert ads into AI chatbot responses without touching the chatbot itself — it works as an add-on that rewrites the final answer to include a sponsored message, without changing how the chatbot was built or trained.
• Compared to just prompting a chatbot to include an ad, PILA produced 34.2% better results; compared to a sampling-based approach, it was 47.3% better; and compared to fine-tuning the base model for ads, it was 7.7% better — measuring a combined score of user satisfaction and ad visibility.
• When PILA was plugged into seven major commercial AI models (like GPT, Claude, Gemini), it improved their combined user-quality and ad-effectiveness scores by 17–18%, without any changes to those models.
• PILA includes a dial (called an 'ad intensity controller') that lets operators choose how aggressive or subtle the ad feels — from a barely-noticeable mention to a more prominent placement — giving publishers a practical way to price different ad tiers.

Marketing implications

• If you manage an AI chatbot or LLM-powered product and want to add sponsored placements, PILA's architecture shows you don't need to rebuild the chatbot — you can add a small rewriting layer on top that handles ads separately. This is worth telling your engineering team about.
• The 'ad intensity dial' concept is directly transferable to pricing strategy: charge more for prominent ad placements, less for subtle mentions. You can build a tiered ad product around this without creating entirely different ad formats.
• Be aware that LLM-native advertising is a fast-moving space — major AI companies including OpenAI are already exploring this. If you buy or sell AI-powered media, start thinking now about how native LLM ads will be disclosed, priced, and measured before industry standards are set.

Apple Podcasts  ·  Spotify  ·  Buzzsprout