Methodology

Company Selection

Our dataset combines companies from three widely cited industry rankings:

• a16z Top 100 Gen AI Apps — 6th Edition
• Forbes AI 50
• LinkedIn Top 50 Most Promising AI Companies — curated by Dinis Guarda

Analysis Framework

Using an LLM, we grouped semantically similar language into ten recurring positioning tropes:

• Trust — safety, reliability, responsibility, ethics
• Human — human potential, human creativity, human-first, humanity
• Change — transform, redefine, reinvent, reshape
• Problem Solving — solve, unlock, remove barriers, eliminate bottlenecks
• Future — frontier, next generation, pioneering, cutting-edge
• Democratization — for everyone, accessible, equal opportunity
• Creativity — imagination, creation, design, storytelling
• Build — build, builders, building
• Ease — simple, intuitive, seamless, frictionless
• Productivity — automate, efficiency, faster

The model evaluates meaning, not just keywords. That means it recognizes paraphrases and synonymous language — not only exact word matches.

Scoring Rules

• Each unique idea is counted once.
• If a phrase could fit multiple tropes, it's assigned to the most specific category to avoid double-counting.

For example: a sentence like "We're building safe AI for everyone" would contribute to the most appropriate bucket(s) without inflating the score by matching every related keyword.

Reproducibility

The underlying dataset — including company names, source URLs, scraped copy, and bucket-level classifications — is available upon request.

Privacy

Anything you analyze here stays private. Your positioning statement is not shared with us or other users. Your copy may be processed by the underlying LLM provider to generate the analysis, subject to that provider's data policies.