The Confidence Layer
AI has a layer of sources it looks to when asked to confirm whether an entity — say, a brand — exists and belongs in a response to a prompt. Let’s call that pool of brands “training data.” And the layer, let’s call that the “Confidence Layer.”
Brands cannot decide if they want to be an entity in an AI’s memory. Of course, all do. The wrinkle is that brands can’t control where they are categorized, how AI describes them, or how much time remains between an AI’s last and next training interval. It could be weeks or months. And brands don’t get an invitation to their AI training party. The brands get embedded and categorized, only to be called up when a possibly relevant prompt swings the spotlight their way.
Brand AI Report is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
If you happen to have rebranded, updated a service, or any other form of normal proactive business activity after AI’s most recent training, AI won’t know it until its next training — which could be tomorrow or next year. For a lucky few (probably well-known) brands, AI will check its Confidence Layer to see if the training data is correct. But for the others in the category, that’s unlikely.
Why does this matter? As people increasingly turn to their AI systems (apps, chatbots, browsers) for product ideas, category searches, etc., the AI doesn’t start from scratch by searching the internet. It’s already done that. In effect, the AI takes the whole pool of brands that could match the criteria in the prompt, and filters that down based on what it finds in its Confidence Layer — a lot of which is online, but also in “knowledge graphs,” structured language that AI trusts, informed by things like Wikidata and Wikipedia. How a brand is described by AI (in response to a prompt) is a very short summary from many sources in the Confidence Layer.
It’s called that because the AI wants to be right. To be right, the AI needs to verify that the brand exists, that what is written about the brand is consistent from source to source, that respected sources have recommended the brand, the brand’s area of expertise and value — and what it is not good at (which also gives AI confidence).
In the next training phase of each AI system, much of the Confidence Layer will update the previous version of the brand embedded in memory, or log it anew.
All of which is to say, brands should pay attention to the Confidence Layer. It’s not glamorous, but it is the main way for brands to influence if and how AI includes them in a response to a user’s prompt. Given that AI only returns a handful of recommendations (from a potential pool of hundreds in a category), brands — especially newer ones — cannot expect their brand guidelines or recent press to get them there.
In this way, AI is like the Hotel California. Brands can “check out” and not know, let alone maintain, their representation in AI’s Confidence Layer, but they can never leave. AI has (or soon will, for new brands) already embedded and categorized them.
Which is why the Confidence Layer matters. Brands can’t control current training data, but they can determine how easily AI can validate it, and what goes into the next training cycle — through accuracy, citations, and consistency from source to source.
Onto Recent AI Headlines for Boards and Brand Leaders:
Google’s AI Commerce Push & Industry Reactions
Inside Google’s push to blend AI chat and online shopping. Marketing Tech News
Shopify, Walmart Endorse Google’s New Open Commerce Protocol. The New Stack
Retailers like Kroger and Lowe’s test AI agents without handing control to Google. AI News
A consumer watchdog issued a warning about Google’s AI agent shopping protocol -- Google says she’s wrong. TechCrunch
Retail Innovation & Enterprise Commerce
How Shopify is bringing agentic AI to enterprise commerce. AI News
The Convergence of Brands and Commerce. www.adweek.com
Retailers bring conversational AI and analytics closer to the user. AI News
Strategic Business Intelligence & Competition
What a New AI Disruption Index Can Tell You About Your Brand’s Future. www.adweek.com
Keeping pace with the competition in the age of AI. The AI Journal
Why Apple chose Google over OpenAI: What enterprise AI buyers can learn from the Gemini deal. AI News
Content Strategy & Methodology
Five LLM Content Strategies Revealed from Top Dev Tool Companies. EveryDeveloper
Reimagining Personalization. bradenkelley.com
The Power of Constraints. Collaborative Fund