Confessing to Claude

Scott Galloway said on his Pivot podcast that he had shared his recent blood work with ChatGPT to get advice. It really struck me. He would never have loaded his blood work onto Google for a perspective, or to Meta for some advice and constantly expresses distrust of both. Why then to an LLM?

Is it true that we are more comfortable sharing our personal secrets with an anthropomorphic chat interface, than we are with a tech platform company?

And the AI companies are happy that we are. Why? Well now that the LLMs have been trained up on known data, and are finding the limits of synthetic data, there is one last main source of data for training and improvement: YOU.

Data privacy has been drilled into me. At Publicis, when we acquired and integrated Epsilon, we were naturally obsessed with Personally Identifiable Information (PII) - keeping it secure, meeting requirements in different parts of the world, carefully matching with client data in a way that preserves the PII they have too. We invested in secure data clean rooms, had endless legal reviews for basic demographic matching, and months negotiating data sharing agreements. Yet here we are, casually feeding our most sensitive information to AI systems run by the same tech companies we've historically viewed with suspicion.

The research reveals something fascinating about human psychology that every CMO needs to understand. People perceive AI fundamentally differently than traditional technology companies, and this perception drives dramatically different privacy behaviours.

Dr. S. Shyam Sundar's research in the Journal of Computer-Mediated Communication identifies what he calls the "machine heuristic" - users perceive machines as "rule-governed, precise, accurate, objective, neutral, infallible, and when entrusted with private information, do not gossip like some humans." This creates a cognitive bias where customers view AI systems as more trustworthy repositories for sensitive information than human-operated business systems.

But it goes deeper. When AI systems use conversational interfaces and demonstrate apparent understanding, users begin applying interpersonal relationship norms typically reserved for human interactions. The turn-taking nature of chat interfaces activates what researchers call Social Presence Theory effects, making users feel they're confiding in a trusted confidant rather than feeding data into a corporate system.

Think about your own behaviour. When you interact with ChatGPT or Claude, do you feel like you're talking to a company, or to an intelligent entity that happens to be powered by a company? The psychological framing matters enormously, and it's reshaping the entire customer relationship landscape.

This creates a fascinating paradox for marketing leaders. Your customers are sharing intimate details with AI systems operated by the same tech companies they refuse to trust with basic demographic data for your campaigns. They're granting calendar access to AI scheduling assistants while blocking your retargeting pixels. They're uploading financial documents to AI tax preparation tools while abandoning shopping carts rather than create accounts with your brand.

Understanding this behaviour isn't just academic - it's becoming the defining factor in customer acquisition and retention strategies.

There is a "privacy paradox" specific to AI suggesting that perceived utility value overrides privacy concerns when mediated through AI chats. But here's the strategic insight that most marketing leaders are missing: customers aren't abandoning privacy concerns - they're applying different risk-benefit calculations to AI interactions. Organisations that demonstrate responsible AI stewardship while delivering clear customer value will establish sustainable competitive advantages before competitors understand what's happening.

Research from Epoch AI projects that technology companies will exhaust publicly available training data between 2026-2032. Elon Musk confirmed in January 2025 that "we've now exhausted basically the cumulative sum of human knowledge in AI training."

This data scarcity has forced AI companies to fundamentally restructure their data acquisition strategies. OpenAI's "Data Partnerships" program now specifically seeks "data that expresses human intention," including user conversations and proprietary datasets. And of course Meta's recent announcement of their data behemoth called "prometheus" is designed to map and predict these intentions based on what they know of all of us.

User interaction data is becoming the primary training source for next-generation AI systems. This creates a feedback loop where customer conversations with AI directly influence the capabilities and biases of future AI systems.

The strategic question for marketing leaders isn't whether customers will continue sharing intimate details with AI systems - they will. The question is whether your brand will position itself as a trusted steward of AI-mediated customer relationships, or watch competitors capture the value of these deeper, more trusting connections.

Those who understand this shift and act decisively will build sustainable competitive advantages that compound as AI becomes more central to customer experience. The window for establishing thought leadership in ethical AI marketing is open, but it won't remain open indefinitely. The data is clear, the psychology is understood, and the competitive landscape is restructuring around these principles.

Thanks for reading, I'm switch back over to Claude, I need his advice on something sensitive.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 22 July 2025 in the Brandflow newsletter on LinkedIn.