How AI Recommends Products to a Narrative

When AI like ChatGPT, Google AI, or Perplexity receives a narrative like this:

“I want to get a new drill for my son – he owns a food truck and needs to drill through a lot of metal. I want it to be compatible with most drill bits – because I noticed some drills were not compatible with others… I would like it to be under $400…”

Here is the process AI tends to use to recommend a product.

————————————

1. Implied Overall Context – Identify

AI will first attempt to understand the overall intent behind the narrative and infer additional context that may be important even if the user did not mention it explicitly.

In this example, AI may infer:

• the user is operating in a mobile environment
• likely working in tight spaces
• drilling through sheet metal
• potentially working around grease, heat, and food-related workflow constraints
• requiring durability and compatibility

LLMs work by recognizing patterns and contextual associations in language to predict useful responses.

——————————–

2. Translate Needs into Technical Requirements

From there, the narrative may be translated into technical requirements such as:

• enough torque to drill through metal consistently
• avoid overheating under repeated use
• compatibility with most drill bits
• avoid:
  • hex-only impact drivers
  • proprietary systems
  • smaller 3/8″ chucks
• look for:
  • 1/2″ metal ratcheting chuck

AI retrieval systems may then identify products matching those requirements through:

• product descriptions
• specifications
• reviews
• forums
• videos
• and other digital narrative environments

———————————

3. Corroborated Truth Validation

For recommendation-based searches, corroboration appears to become increasingly important.

AI retrieval systems appear to strengthen recommendation confidence through repeated contextual associations across digital narratives.

Forums like Reddit often contain valuable experiential feedback where people describe:

• similar workflows
• operational problems
• product tradeoffs
• failures
• and successful solutions

This allows AI systems to potentially identify contradictions between:
“This product can drill through metal”
vs.
“I used this drill for metal work and it overheated quickly.”

Over time, AI retrieval systems may become increasingly capable of distinguishing between:

• marketing claims
• repeated semantic positioning
• and real-world operational experiences.

AI systems appear to associate products and brands with recurring narratives such as:

• “good for food truck”
• “overheats quickly”
• “not ideal for metal”
• “compatible with most bits”

These repeated contextual associations across reviews, forums, videos, articles, and product pages contribute to a product’s broader digital semantic footprint.

————————-

4. Tradeoffs 

AI may then make tradeoffs between:

• price
• performance
• constraints
• durability
• compatibility
• and workflow fit

before generating a recommendation.

Narr Theory refers to this emerging behavior as Narrative Product Discovery.

It can also be described as:

• Conversational Product Discovery
• Semantic Product Discovery
• Narrative Retrieval

The overall concept is relatively simple.

What makes AI retrieval systems powerful is the scale and speed at which they interpret contextual associations across digital narratives.