Most viewers mistake AI-edited video for human editing in new study

A 315-person study from Editvideo.io found that human-edited video performed better on engagement and trust, but a majority of viewers still misidentified the AI-edited version as human-made. The results suggest AI video editing is getting harder for everyday viewers to spot, even as human editing still wins on the metrics that matter most to creators.

Why it matters: - The findings point to a shift in how viewers perceive edited video. - Human editing still outperformed AI on engagement and trust, but AI output was convincing enough to fool most viewers. - The study suggests creators may need to think less about whether audiences can spot AI and more about how editing choices affect attention and credibility.

What happened: - Editvideo.io released an independent study testing whether everyday viewers could tell the difference between a video edited by a professional human editor and one edited entirely by AI. - The study surveyed 315 US-based respondents. - Each respondent watched the same raw footage edited two ways: one version by a professional human editor on Editvideo.io's team and one version by Descript's Underlord AI co-editor using a single prompt with no manual changes afterward. - Respondents were asked which version was more engaging, which felt more trustworthy, and which one they believed was edited by a human.

The details: - Human-edited video won decisively on the core viewer-response questions. - 78.7% of respondents said the human edit was more engaging. - 63.2% said the human edit felt more trustworthy and made them more likely to subscribe. - When asked to identify the human-edited version, 53.3% of respondents chose the wrong one and believed the AI-edited video was human-made. - The strongest pattern in the responses was an editing-style assumption. - Many respondents assumed simpler, plainer editing was human-made and more polished, graphics-heavy editing was AI-made. - In this study, that assumption did not hold. - The human editor used branded motion graphics and highlighted captions. - The AI-edited version stayed deliberately minimal. - Accuracy varied by education and income. - Respondents with a graduate degree identified the human edit correctly 34.8% of the time. - Respondents with a high school education were correct 45.5% of the time. - Accuracy declined as household income rose, from 51.8% among respondents earning under $35,000 to 40.0% among those earning $150,000 or more. - Editvideo.io said it designed the study to be transparent about its commercial interest in the outcome. - The company published the full methodology, the AI prompt used, and the complete dataset alongside the results. - The company also disclosed an early draft that dropped the video's closing call to action and had to be corrected. - The full report, including both original videos side by side, demographic breakdowns, and respondent quotes, is available on the company's website.

Between the lines: - The study suggests viewers may equate polish with automation, even when the opposite is true. - That matters for creators and brands because the visual style of an edit may shape trust as much as the editing tool used. - Editvideo.io also appears to be using the study as both a market signal and a credibility test for its human-editing business.

What's next: - Editvideo.io is making the methodology and supporting materials available for public review. - The company is likely to use the findings to argue that human editors still have an edge where audience engagement and trust are concerned. - The broader question is whether AI editing will keep improving fast enough to blur the distinction even further.