What Is AGI? A Precise Definition for 2026
Artificial general intelligence (AGI) defined: a system that matches or exceeds competent human adults across the cognitive tasks required for most economically valuable work, without being retrained for each one.
Executive summary
Artificial general intelligence (AGI) is the hypothesised milestone at which a single integrated machine system can learn, reason, and act across the breadth of human cognitive work. It is defined by generality and transfer — not by raw benchmark scores on any one task. As of 2026, frontier systems show emerging generality but do not yet meet the standard most researchers use for AGI.
Key concepts
- Generality across domains
- Transfer learning between tasks
- Autonomy under novel conditions
- Hypothesised research milestone
- Distinct from consciousness
A working definition
A practical definition used by most laboratories in 2026: an AGI is a single integrated system that can match or exceed competent human adults on the cognitive work required for most economically valuable jobs, across novel domains, without being retrained for each one.
Three properties carry the weight in that sentence. Generality — the same system handles many task families. Transfer — skill learned in one domain helps in another. Robustness — performance does not collapse when the problem shifts slightly from training data.
Why the term exists
Early AI research in the 1950s and 1960s assumed general intelligence was the natural goal. When that proved harder than expected, the field split into narrow subfields — vision, speech, planning, robotics — each making progress in isolation. By the 2000s researchers including Ben Goertzel and Shane Legg reintroduced the phrase artificial general intelligence to distinguish the original ambition from the narrow systems that had become the commercial mainstream.
Where today's systems stand
Google DeepMind's 2024 Levels of AGI framework places today's strongest chatbots at Level 1 — Emerging AGI: equal to or somewhat better than an unskilled adult across a wide range of non-physical tasks. Competent AGI (matching the 50th percentile of skilled adults) and Expert AGI (90th percentile) remain unreached as integrated, persistent systems.
Frontier reasoning models — OpenAI's o-series, Anthropic's Claude with extended thinking, Google DeepMind's Gemini, DeepSeek-R1 — score at or above expert human level on isolated benchmarks. Generality, however, is measured by the breadth of tasks handled without retraining, not the height of any single score.
What AGI is not
AGI is not consciousness. Generality of cognition and subjective experience are separate questions and the second is not required for the first.
AGI is not a specific product. It will likely arrive as a continuum of increasingly general systems rather than as a single launch.
AGI is not inevitable. It is a research goal with significant open problems in continual learning, long-horizon planning, memory, embodiment, and alignment.
Key takeaways
- 01AGI is defined by generality and transfer, not by score on any single benchmark.
- 02Today's frontier systems sit at the Emerging AGI level — strong but narrow in practice.
- 03Consciousness is a separate question from general intelligence.
- 04The path to AGI is plural: many capability axes must mature together.
Frequently asked questions
Is ChatGPT AGI?
No. ChatGPT and other frontier chatbots show flashes of general reasoning but lack continuous learning, persistent memory, and reliable long-horizon planning. They sit at Level 1 (Emerging AGI) in DeepMind's framework.
When will AGI arrive?
Serious researchers disagree, with credible estimates ranging from this decade to many decades away. The 2025 International AI Safety Report concluded that timing is genuinely uncertain and that no single forecast should be trusted.
Is AGI the same as superintelligence?
No. AGI matches competent humans across cognitive tasks. Artificial superintelligence (ASI) substantially exceeds the best humans across essentially every domain — a regime that may follow AGI but is not the same thing.