Archive · 30 Aug 2026
Why Puzzles Endure While AI Still Struggles to Create Them
Puzzles have been entertaining people for hundreds of years.
The first jigsaw-like puzzles were maps cut into pieces for students.
Later, newspapers popularized crosswords, riddles and other word games.
People enjoy puzzles because solving them can bring a rewarding “aha” feeling.
Puzzle makers try to make clues challenging but not impossible.
They also use jokes, wordplay and references from different cultures and generations.
Some puzzle publishers express particular political or social viewpoints, while others aim to stay neutral.
Computers can solve many puzzles and even build some crosswords.
However, people say AI still does not understand human humor and clever cryptic clues as well as skilled puzzle setters.
Puzzles evolved from John Spilsbury’s 1770s dissected maps into popular word and logic games.
The global word-games and logic-puzzle market is valued at $9.4 billion and projected to reach $20.1 billion.
Setters say a good puzzle balances difficulty with fairness, humor, originality and a satisfying solution.
Growing international audiences have encouraged puzzles to include broader cultural, generational and gender perspectives.
AI can solve and construct some crosswords, but experts say it still struggles with strong cryptic clues and human humor.
- Who
- Puzzle solvers and setters, including Will Shortz, David Astle, Mick Hodgkin, Ben Tausig, Patti Varol, Sowmya Ramkumar, Tony Sebastian and John Henderson.
- What
- The article examines why people enjoy puzzles, how puzzles are created and why AI has not yet matched humans at setting high-quality cryptic clues.
- When
- The history discussed begins in the 1770s and continues through developments and examples cited from 1913, 1940s, 2021, 2024 and mid-2026.
- Where
- The story covers puzzle publications and communities in the United States, United Kingdom, Australia, India and other countries.
- Why
- Puzzles offer challenge, curiosity, humor and the satisfaction of reaching an answer, while AI lacks some human understanding needed to create especially effective clues.
Human-Centered Puzzle Making
Technology-Driven Puzzle Solving
Can AI create good puzzles?
Human-Centered Puzzle Making
Setters argue that strong puzzles require human judgment about humor, assumptions, fairness, cultural references and emotional payoff; AI still struggles especially with cryptic clues.
Technology-Driven Puzzle Solving
AI can already construct full New York Times-style crosswords and automated systems can solve difficult puzzles, showing that machines can perform important parts of the puzzle-making process.
Political and social viewpoints
Human-Centered Puzzle Making
AVCX embraces the setter’s voice and describes its politics as liberal; Patti Varol also says puzzles should represent LGBTQ people and a broad spectrum of human life.
Technology-Driven Puzzle Solving
Legacy newsroom puzzles often aim for a uniform, objective or apolitical voice, and Varol says her publication seeks to remain non-controversial even while acknowledging representation.
Local references versus global access
Human-Centered Puzzle Making
International audiences benefit when puzzles broaden their references across cultures, ages and genders, including Indian food, places and personalities.
Technology-Driven Puzzle Solving
Highly local clues and references can remain difficult for people outside the original community, and broader puzzles may still rely on stereotypes or unfamiliar cultural knowledge.
- Early puzzle origin
- In the 1770s, John Spilsbury mounted a world map on wood and cut it into pieces as an educational exercise.
- First crossword
- Arthur Wynne created a diamond-shaped crossword called “Word Cross” for The New York World in 1913.
- Market size
- The word-games and logic-puzzle market is valued at $9.4 billion, according to the cited Dataintelo outlook.
- Projected market
- The same outlook projects the market will grow to $20.1 billion over the next decade.
- New York Times Games
- The New York Times Games app reportedly generated about $11.2 million in worldwide in-app purchase revenue in the second quarter of 2024.
- International subscribers
- More than three million of the New York Times Games’ 13.35 million total subscribers were outside the United States as of mid-2026.
- AI performance
- Crossword Genius solved five of six puzzles at the 2024 Times Crossword Championship, but did not solve John Henderson’s cryptic puzzle.