Marketing leaders have worked out that AI carries an environmental cost. Very few have worked out how big it is.
New benchmark research surveying 100 US and 100 UK senior marketers found 88% believe artificial intelligence is increasing their organisation’s carbon footprint. Yet only 36% have comprehensively measured that impact, and 8% have not measured it at all.
That is a belief-to-evidence gap wide enough to undermine every sustainability claim sitting alongside it.
The Numbers Behind the Dissonance
The headline findings paint a consistent picture:
- 88% of senior marketers say AI is increasing emissions
- 42% believe the increase is significant
- Only 36% have fully measured their AI environmental impact
- 8% have carried out no measurement whatsoever
Regional split is notable. 51% of US marketers perceive a significant emissions uptick from AI, against 32% of UK marketers. Whether that reflects genuinely different infrastructure exposure or simply different levels of scrutiny is unresolved.
Cost Is Climbing Too
Emissions are not the only pressure point. 88% of respondents say AI is pushing operational costs up, with 35% describing the increase as substantial.
So the technology is simultaneously raising the energy bill, the cloud bill, and the reputational exposure. Most teams are tracking only one of those three.
Progress Elsewhere Makes the Gap Worse
The awkward part is that sustainability programmes are otherwise improving. Only 17% of marketers now cite budget as a concern in this area, down 20 percentage points on 2025. Fewer than a quarter see internal alignment as a key obstacle. And 85% report moderate to significant progress in cutting emissions overall.
In other words, the funding is there, the buy-in is there, and the trajectory is positive. AI is the one line item nobody has instrumented.
The research frames this as a governance problem rather than a willingness problem. Marketing teams adopted AI tooling at speed, often outside procurement, and the environmental accounting never caught up.
Why This Matters for Brands Right Now
Data centre energy consumption has moved from a technical footnote to a live consumer concern. Public sentiment towards AI infrastructure has hardened noticeably over the past year, and investor scepticism around AI valuations is sharpening in parallel.
Any brand publishing sustainability targets while running unmeasured AI workloads is carrying a credibility risk it has not priced in. The claim is only as strong as the measurement behind it.
What Marketing Teams Should Do Next
Three practical moves:
- Inventory your AI stack. List every model, platform, and automation in active use, including tools adopted informally by individual teams.
- Request energy and emissions data from vendors. Several major providers now publish per-query or per-workload figures. Ask for them contractually.
- Fold AI into existing carbon reporting. Treat it as a category within Scope 3 rather than a separate exercise nobody owns.
The organisations that measure first will be the ones able to make defensible claims later. Everyone else is guessing in public.
