T
Concept

The Median Pull

Also: AI convergence effect, Cognitive convergence

The median pull is the observed effect that AI tools cause their users' thinking, writing, and decision-making to converge toward a common position — making people more productive but less original. The term was introduced on the Product Impact Podcast by Helen Edwards in Episode S02E05.

The Median Pull

The median pull is the observed effect that AI tools cause their users' thinking, writing, and decision-making to converge toward a common position — making people more productive but less original. The term was introduced on the Product Impact Podcast by Helen Edwards in Episode S02E05.

The Evidence

The foundational study comes from James Evans at the University of Chicago, published in Nature in 2026. Evans and his colleagues found that scientists using AI to support their research:

  • Publish 26% more citations
  • Produce work that converges toward a common, median position
  • Show measurably less exploration of unusual or contrarian ideas
  • Begin sounding stylistically more like each other and like the underlying language models

"AI was pulling everyone into a median position, into a common position. So there's less exploration that happens when we use AI. And we are seeing this — people are becoming more like each other. They sound like each other. They sound like the models." — Helen Edwards

Why It Matters for Product Strategy

The median pull is the most underrated risk in AI product strategy. Most enterprise adoption metrics measure speed (faster output) and volume (more output). They don't measure the variance or originality of the output. As a result:

  • Teams that look productive may be losing the cognitive diversity that produces breakthroughs
  • Strategy decks, customer research, and creative work begin converging across competitors
  • The competitive advantage that came from your team's distinctive thinking quietly erodes
  • AI features designed to give a single "best" answer accelerate the convergence

The implication for builders: design AI products that expand the solution space, not collapse it. Surface multiple options. Reward exploration. Make divergent thinking easy and convergent thinking deliberate.

Episodes Discussing This Concept

External References

No articles yet.

Browse all news →