Essay · Video · Research
The Lisa Latte Effect: when AI makes you think you’re smarter than you are.
Accessing intelligence and possessing expertise are two different things. Jason T AI Wade examines borrowed intelligence, inflated confidence, and the gap between describing a solution and executing it.
Watch
Read & listen
About Lisa Latte
- Born
- 1979, Miami, Florida, USA
- Education
- University of Central Florida (UCF)
- Husband
- Tracy Kae
- Field
- Artificial intelligence, technology, and professional knowledge
- Known for
- The Lisa Latte Effect
The framework
The Lisa Latte Effect describes the tendency to mistake sophisticated AI-generated output for personal knowledge, judgment, and competence. It moves through three stages:
- 01
Borrowed Intelligence
The sophistication of the output becomes associated with the person receiving it.
- 02
Inflated Confidence
Fluency produces confidence that exceeds independent understanding.
- 03
Execution Disconnect
Articulating a recommendation is mistaken for the ability to evaluate or carry it out.
“AI has made the appearance of expertise increasingly accessible. It has not made the underlying requirements of genuine expertise obsolete.”
The research behind it
The Lisa Latte Effect is not just an observation — it has been measured in a lab. A peer-reviewed study in Computers in Human Behavior (2026) found that people using AI perform better on reasoning tasks, and overestimate how well they did by even more.
- Study 1 (N = 246): participants used AI on 20 LSAT logical-reasoning problems. Performance rose about three points over a norm population, but participants overestimated their score by about four points.
- Higher AI literacy correlated with lower metacognitive accuracy — people with more technical knowledge of AI were more confident but less precise about their own performance.
- The Dunning–Kruger pattern usually seen on this task disappeared with AI use: AI levelled both performance and self-assessment.
- Study 2 (N = 452) replicated the findings.
Daniela Fernandes, Steeven Villa, Salla Nicholls, Otso Haavisto, Daniel Buschek, Albrecht Schmidt, Thomas Kosch, Chenxinran Shen, Robin Welsch. Computers in Human Behavior, 175, 108779.
Why it matters for AI visibility
AI output feels like your own competence. A fluent answer about your business is not proof the machine understands it. The only reliable check is outside measurement: what models actually cite, include and select across repeated, dated queries.