If you build or buy video tooling, you have probably seen two confident numbers about AI adoption that cannot both describe the same market. One survey says three in four video marketers use AI. Another says fewer than one in five. Both are real and both are published. The spread says more about how the questions were asked than about how many teams actually use AI.
At Mark Studios we trace popular video statistics back to the organizations that produced them. Here is what the AI video numbers measure, and a short checklist for reading the next one.
The Same Publisher, Three Different Answers
Wyzowl, which sells explainer videos, has asked video marketers nearly the same question for three years. In its 2024 report, 75% said they had used AI tools to create or edit marketing videos. In the 2025 edition the share was 51%. In 2026 it was 63%.
Read as a trend, that says adoption fell by a third and then recovered. More likely, differences in the samples and question wording contributed to the changes: 967 respondents in late 2023, 205 in late 2024 and 266 in late 2025, and in all three years the published sample combined marketers and consumers. With samples that small and that different, a year-over-year change tells you very little.
Different Questions, Different Universes
Widen the lens and the range gets wider. Wistia, which sells video hosting and AI tools, reported that 18% of surveyed professionals were using AI to make videos in 2024. In a separate Wistia survey of 503 marketers who work with video, fielded from September 16 to October 1, 2025, 99% said they had experimented with AI.
Both numbers come from the same company. The difference is the question. “Are you using AI to make videos?” and “Have you experimented with AI?” describe different behaviors, and only one of them says much about what a team will pay for next quarter.
A useful rule: the looser the verb (experimented, tried, explored), the higher the number.
What the Government Surveys Say
Government surveys don’t sell anything. The U.S. Census Bureau’s Business Trends and Outlook Survey found that 24.6% of U.S. employer businesses used AI in the October 5 to 18, 2026 collection period, up from 19.8% in the period ending May 3, 2026, with the same question in both. In the Information sector, which includes the motion picture and video industries, the share was 46.4%.
The Census survey doesn’t ask about video. European and UK statisticians do. In 2025, 9.55% of EU enterprises with 10 or more employees used AI that generates pictures, videos or sound, according to Eurostat. In the UK, the Office for National Statistics found that 16% of businesses with 10 or more employees used AI for visual content creation in June 2026.
Those shares are far lower than the vendor surveys for a simple reason: they count businesses of every kind, not people who already work with video.
Usage Counts Are Not Usage
The other kind of number you will see is a company’s own count. Kuaishou said its Kling model had generated more than 600 million videos by December 2025. Google said more than 275 million videos had been generated in Flow, its Veo-based filmmaking tool, by October 2025. Adobe reported 29 billion Firefly generations by September 2025 and has never published a count of videos alone; its unit moved from images to generations to assets along the way.
These counts are real signals of activity, but none of them says how many of those videos were published or watched. A generation is a button press, not a deliverable.
The Forecast That Became a Fact
Then there is the number that was never a measurement. “90% of online content will be AI-generated by 2026” circulated for years with Europol’s name on it. Traced back, it was a single prediction by the chief executive of an AI video company, made about video rather than all content. Europol published a revised edition of the report in January 2024 without the sentence and said a statement from an inaccurate source had been removed.
A Checklist for the Next AI Video Statistic
Before a number goes into a roadmap, a budget or a slide, ask:
- Who ran the survey, and what do they sell? Many adoption surveys in this space come from companies that sell video production, hosting or AI tools, so it’s important to check whether their samples are representative.
- What was the exact question? “Used to create or edit,” “experimented with” and “uses in the workflow” are three different behaviors.
- Who was asked? A vendor’s customers, marketers who work with video, or all businesses?
- How many people, and when? A sample of 205 collected in one season is not a trend line.
- Is it a count or a share? Cumulative counts only go up and say nothing about work that ships.
- Is it a measurement or a forecast? If the earliest version you can find is a prediction, treat it as one.
None of this means AI is not changing video work. It means the honest answer to “how many teams use AI for video?” is a range, and where a team sits in that range depends on what you count as using it.
The surveys, government figures and company counts cited here, with their samples and dates, are collected at https://www.markstudios.com/research/ai-video-editing-statistics#how-many-use-ai-for-video
About the author: Mark Santos is the founder of Mark Studios, a video studio in Palm Beach County, Florida. He has edited video since he was 10, starting on his own YouTube channel, and began selling his editing on Fiverr in 2020. Mark Studios has since completed 8,000+ orders for 4,500+ clients, and it publishes research that traces popular video statistics back to their original sources.
