\n<\/p>\n
I started noodling on this topic with the premise that the AI genie is out of its bottle and it ain\u2019t going back in.* So, given the concerns around AI \u2013 jobs, IP rights, <\/span>slop<\/span><\/a>, poor quality, users being lazy, let alone environmental and community <\/span>concerns around data centers<\/span><\/a> \u2013 it got me thinking:\u00a0 How can the ad industry sustainably use AI in marketing?\u00a0<\/span><\/p>\n Let\u2019s find out.<\/span><\/p>\n Quigley-Simpson co-founder Gerald Bagg<\/b> notes that, for the most part, the environmental impact of AI is not a \u201croutine client question\u201d yet.\u00a0<\/span><\/p>\n \u201cMost organizations remain more focused on capability, speed, cost, governance and accuracy. But environmental impact will become harder to ignore as AI moves from occasional experimentation to continuous production at scale,\u201d Bagg says.<\/span><\/p>\n Neil Clark, chief sustainability officer of MSQ Partners, parent of Big Spaceship<\/b>, also says it\u2019s not a top question from clients, but he expects more clients will start asking it.<\/span><\/p>\n \u201cIt\u2019s a significant topic internally,\u201d Clark adds. \u201cOur people raise it regularly, they want to act on it and they push us to do better. That matters because our clients\u2019 teams are made up of the same kind of people. If our own employees are asking these questions, theirs will be too.\u201d<\/span><\/p>\n While there is some data online about the <\/span>energy usage required for AI prompts and creations<\/span><\/a>, many of the big players do not disclose such details or are current with it.<\/span><\/p>\n Let\u2019s start small: A <\/span>typical<\/span><\/i> interaction on OpenAI\u2019s ChatGPT uses about 0.3 watt-hours of electricity. Since a standard LED light bulb uses 10 watts, that AI query equates to about 2 minutes of powering the bulb.<\/span><\/p>\n How about images? A generative AI image takes approximately 1.2 watt hours, or the same energy as having an LED light bulb on for 7 minutes.\u00a0<\/span><\/p>\n Video? That can take 30 to 950 watt hours, depending on quality. Looking at the higher end, creating a generative AI video can equate to leaving that light bulb on for four days.<\/span><\/p>\n If we compound the above information across the entire industry, that\u2019s A LOT of light bulbs with the switch on.<\/span><\/p>\n With that being said, the energy to process prompts is becoming more efficient. <\/span>Google says that<\/span><\/a>, in May 2024, the medium Gemini prompt used 33 times more energy than it did a year later. The tech giant says software optimization, including improvements in models, contributed to the increase in energy efficiency.\u00a0<\/span><\/p>\n Note: Scroll to the bottom to see how Ad Net Zero\u2019s Global Media Sustainability Framework and AdGreen help measure the ad industry\u2019s AI-related emissions.<\/span><\/i><\/p>\n A May <\/span>Bentley University-Gallup Business in Society Survey<\/span><\/a> showed that 73% of US adults don\u2019t trust businesses (either not much or not at all) to responsibly use AI. While that\u2019s an improvement from 2023 (78%), mistrust this year is up when compared to 2025 (69%).<\/span><\/p>\n Let\u2019s get closer to marketing.\u00a0<\/span><\/p>\n One in 5 millennials has a <\/span>negative view<\/span><\/a> of AI\u2019s use in marketing, which soars to nearly 2 in 5 when you ask Generation Z, per an IAB study released earlier in 2026.<\/span><\/p>\n When seeking an emotional connection, generative AI may not be the way to go. Kantar found that, while generative AI-created ads can spark emotion in consumers, that emotion is often <\/span>not what was intended<\/span><\/a> and is negative.\u00a0<\/span><\/p>\n How about the sought-after trust factor with consumers? <\/span>Klaviyo found that<\/span><\/a> 31% of global consumers put <\/span>less trust<\/span><\/i> in a brand that uses generative AI, though 61% were neutral.<\/span><\/p>\n Getting back to the premise, there are ways today to use AI in marketing that minimize its environmental impact.\u00a0<\/span><\/p>\n \u201cThe biggest step that you can take is the call you never make,\u201d MSQ\u2019s Clark says. \u201cPlenty of tasks don\u2019t need an AI at all, and simply asking \u2018does this actually need a model?\u2019 avoids a surprising amount.\u201d\u00a0<\/span><\/p>\n Clark adds that, \u201cWhen something does need AI, the next question is whether it\u2019s going to the right model, which isn\u2019t always the smallest one. A model that\u2019s too weak for the job just sends you round in circles regenerating answers, and that wastes far more [energy] than choosing the capable option the first time. <\/span><\/p>\n \u201cEqually, reaching for the most powerful model for a task [that] a lighter one could handle is poor practice.\u201d<\/span><\/p>\n Quigley-Simpson\u2019s Bagg echoes that approach, noting that strategic thought is needed when using AI tools.<\/span><\/p>\n \u201cAvoid endless generation simply to produce more alternatives. Reuse approved outputs rather than recreating them. \u2026 Reserve highly computational tools, particularly image and video generation, for situations in which they create genuine value,\u201d Bagg says.<\/span><\/p>\n Bagg also says agencies and clients need to expect more from their partners.<\/span><\/p>\n \u201cOrganizations should also ask their technology partners for greater transparency about energy use, data-center efficiency, model size and the source of the electricity supporting their systems. Marketers cannot manage what providers do not disclose,\u201d he says.<\/span><\/p>\n Clark asserts that using task-specific AI tools may be more efficient than a \u201cgeneral purpose giant.\u201d\u00a0<\/span><\/p>\n His agency took this approach with MSQ Assist, an internal AI platform. MSQ Assist \u201croutes each task to the most appropriate model rather than defaulting to the biggest. Assist is also built on our own knowledge of our clients and is heavily templated, so it steers people towards focused, well-structured prompts and processes rather than a blank box\/page every time,\u201d Clark explains.<\/span><\/p>\n Today and going forward, Clark says that there are two key reasons for thinking about sustainable use of AI in advertising.<\/span><\/p>\n The first is simple: \u201cAI has a real and fast-growing energy and water footprint, and marketing is adopting it at pace, so the totals mount up quickly,\u201d Clark says.<\/span><\/p>\n The second leans into the important role the marketing industry plays in society.<\/span><\/p>\n \u201cMarketing doesn\u2019t only consume energy, it shapes what people see as normal and desirable. If we use AI to manufacture over consumption, or we model wasteful, throwaway AI habits ourselves, we scale that behavior across millions of people. We could just as easily do the opposite.<\/span><\/p>\n \u201cTherefore, AI\u2019s impact isn\u2019t only about our own footprint, it\u2019s about whether an industry with this much influence chooses to use one of the most powerful tools it has ever had thoughtfully, or not,\u201d Clark says.<\/span><\/p>\n The use of AI in marketing, Bagg says, also plays into a much broader question that agencies always need to ask\u201d<\/span><\/p>\n \u201cDid this work need to be created, and did it produce enough value to justify the resources consumed?<\/span><\/p>\n \u201cThe more consequential and potentially controversial argument is that responsible AI use may sometimes mean deciding not to use AI at all. That challenges the prevailing assumption that greater AI adoption is inherently progress, and it is a sufficiently distinct position to warrant attention,\u201d Bagg says.<\/span><\/p>\n \u00a0<\/p>\n As Clark and Bagg point out, the question about sustainable AI in marketing is not so much about the use of AI, it goes to the foundational questions that must be asked at the start of projects: What do you want to accomplish and what\u2019s the best way to do it. And, of course, you want to do it as effectively and efficiently as <\/span>possible.<\/span><\/p>\n \u00a0<\/p>\n \u00a0<\/p>\n * As Generation Z gets more ingrained in the workforce, it will have a <\/span>say in AI usage<\/span><\/a>. Also, if you disagree or have thoughts on the premise, <\/span>email me<\/span><\/a>.<\/em><\/p>\n \u00a0<\/p>\n There are three ways that Ad Net Zero\u2019s <\/span>Global Media Sustainability Framework<\/span><\/a> captures emissions from AI use: Unit operations formulae, \u201casset manipulation\u201d unit operations and \u201ccorporate overhead emissions\u201d allocation.\u00a0<\/span><\/p>\n Ad Net Zero Global Director Jo Fenn says the framework will likely expand from creative workflow to include media delivery, optimization and other ways, including AI as a media channel.\u00a0<\/span><\/p>\nWhat is AI\u2019s environmental impact?<\/span><\/h2>\n
Do consumers want AI in their advertising?<\/span><\/h2>\n
Sustainable use of AI in marketing<\/span><\/h2>\n
Sustainable considerations for adland<\/span><\/h2>\n
AI impact: Ad industry resources<\/span><\/h2>\n