Same Face, Same Brand: Recurring Features in Marketing and the Consistency Gap in Generative AI Video

Marketing works through memory. Repeated exposure makes a stimulus easier to process and better liked, and brands that keep their distinctive assets, characters and creative style consistent over several years report more very large brand and business effects. Generative AI video makes it cheap to produce many short videos, but by default it keeps nothing recurring: every generation is a fresh sample. This paper connects the two literatures. From consumer psychology and advertising effectiveness research we summarise why recurring features matter, including for business buyers who are mostly out of the market at any given moment. From computer vision we summarise evidence that keeping the same character, subject, text and logo across separate generations is a recognised open problem, visible in benchmarks and error audits of 2024 and 2025 models. From both we derive five design principles for brand consistent AI video: store the brand once outside the prompt, anchor each recurring character to one reference portrait, make constraints explicit, approve a still before paying for motion, and keep each generated unit short. We illustrate the principles with GetReelCom, a web application that turns a company website into short video scenes, described from its public documentation. The paper reports no evaluation of that product; instead it proposes an evaluation protocol and a research agenda. Competing interests: the author is a co-founder of GetReelCom, the application used as the illustrative case. About the author: Amber Kesselaer is a growth consultant and an expert in marketing and AI video generation.