A new benchmark study puts a number on an assumption marketing leaders have been carrying since generative AI became a daily tool: that all the images, copy variants, and personalization tests being generated are adding an environmental cost nobody in the marketing organization is actually tracking. 51toCarbonZero’s 2026 “Zeroed In: The Brand Marketing Pulse” report, based on a June 2026 survey of 200 senior marketing leaders across the U.S. and U.K. conducted with Censuswide, finds that 88% believe their AI use is increasing emissions, and 42% believe the increase is significant. Only 36% have comprehensively measured that impact, and 8% have not measured it at all.
The gap matters because AI has moved from a pilot project to production infrastructure inside most marketing teams over the past two years, the same window in which marketers have been absorbing a growing hidden labor cost from reviewing and correcting AI output. The environmental line item is following the same pattern: real, growing, and largely unmeasured, which makes it difficult for CMOs to answer the resource-cost question sustainability and finance teams are starting to ask about AI spending. Eighty-eight percent of respondents also said AI is pushing up their operational costs directly, with 35% reporting substantial increases.
The more interesting number in the data is the direction of travel: concern about sustainability budgets has fallen 20 points from last year, even as 85% of marketers report meaningful progress cutting emissions overall and the perceived AI emissions problem has grown. Marketing leaders are more confident in their broader climate story than they were a year ago, which means the AI footprint risks becoming the one input nobody is watching, precisely because the rest of the sustainability picture looks like it is improving. As boards begin asking for the same governance rigor around AI’s resource use that they are already demanding around what AI agents are allowed to do unsupervised, measurement, not intent, is what will separate teams that can answer the question from teams that cannot.
Source: 51toCarbonZero
