New York Beats Paris and London Combined in AI Prompts
Quick answer
New York appeared in prompts from 2,421 observed Magic Hour creators, compared with 2,251 creators mentioning Paris or London after deduplication. That gives New York a 170-creator lead, or 7.6%.
New York appeared in prompts from 2,421 observed Magic Hour creators, more than the deduplicated total of 2,251 creators who mentioned Paris or London.
That gives New York a 170-creator lead, or 7.6%, over the two European capitals combined. New York also led each city across image, video and audio projects and in every mature quarter reviewed.
The finding describes where Magic Hour creators set their AI-generated scenes. It does not measure tourism demand or the preferences of the general public.
The city ranking
New York appeared in prompts from 2,421 distinct creators. London appeared for 1,183 creators and Paris for 1,180. Because some creators mentioned both European cities, their combined comparison was deduplicated at the creator level, producing 2,251 creators rather than simply adding the two totals.
Why New York keeps winning
The data cannot establish why creators selected a city. New York can function as a shorthand for ambition, fashion, romance, nightlife, finance, density or cinematic scale. Those are possible interpretations, not measured motivations.
Create a city scene
Describe the city, subject, time of day, and one camera move. Generate a short draft, then check landmarks, signs, and continuity.
Try AI Video GeneratorMethodology
The analysis covers completed, enabled, nondeleted Magic Hour image, video and audio projects from June 16, 2023 through September 19, 2026. It counts distinct observed creators whose stored prompt text explicitly mentioned New York, NYC, Manhattan, Paris, the Eiffel Tower, London or Big Ben.
Templates, logged automated free-tool tasks, blank prompts, exact prompts reused by five or more creators in the same month, adult content, named political and celebrity subjects, and prominent copyrighted-character contexts were excluded. The Paris and London comparison was deduplicated by creator. Results were checked across media, workflows, time periods and heavy-user concentration.
The 2x estimate for unlogged free-tool usage is not included in observed counts or percentages. Magic Hour users are not representative of everyone, and the findings should not be generalized to the public.
No raw prompts, personal information, project identifiers or user media are published.
