The Problem With Carbon Offsets That Climate Scientists Keep Raising

Carbon offsets let companies pay for emissions reductions elsewhere instead of cutting their own, but a growing body of peer-reviewed research shows most offset credits overstate their climate benefit, sometimes by a factor of five to ten, because of weak additionality standards, inflated baselines, and forest projects that later burn, get logged, or were never at real risk of disappearing.

That is no longer a fringe accusation. It is the conclusion of a systematic review published in the Annual Review of Environment and Resources, which found that the most widely used offset programs continue to greatly overestimate their probable climate impact, often by a factor of five to ten or more.

The review, led by physicist and climate policy researcher Joseph Romm at the University of Pennsylvania, argues the failures are not isolated to a handful of bad actors. Credit quality has remained a problem since the inception of carbon credits, despite repeated efforts to address the core challenges of additionality, leakage, double counting, environmental injustice, verification, and permanence; combined, these issues have led many researchers to conclude that overcrediting in carbon offsets is an intractable problem.

For anyone buying credits, advising a company on its net zero strategy, or simply trying to understand why the phrase carbon neutral is disappearing from advertising, the practical question is not whether offsets have flaws. It is which flaws are structural, which are fixable, and what that means for how offsets should actually be used going forward.

Why “Additionality” Is the Word That Sinks Most Offset Claims

Every credible carbon credit rests on one counterfactual claim: this emissions reduction would not have happened anyway. Carbon market practitioners call this additionality, and it is the single hardest thing to prove and the easiest thing to fake.

A wind farm that was already financially viable without carbon revenue is not additional, even if it displaces fossil generation. A forest that was never going to be logged is not protected by a REDD+ credit, even if the paperwork says otherwise. The distinction sounds abstract until it is applied to real transactions, and that is where the criticism has landed hardest.

The most consequential test case involves Verra, the world’s largest carbon credit registry, and its REDD+ rainforest protection projects. A nine-month investigation by the Guardian, Die Zeit, and the nonprofit newsroom SourceMaterial, published in January 2023, concluded that journalists who analysed the findings of three scientific studies using satellite imagery to check the results of forest offsetting projects found that in 32 projects where it was possible to compare Verra’s claims with the study findings, baseline scenarios of forest loss appeared to be overstated by about 400 percent.

The threat to forests had been overstated by roughly 400 percent on average, drawing on an earlier 2022 University of Cambridge study. Companies including Gucci, Salesforce, BHP, Shell, and easyJet had purchased offsets tied to those inflated baselines.

Verra disputed the methodology sharply. The registry said the claims were based on studies using synthetic controls that do not account for project-specific factors that cause deforestation, and argued these studies massively miscalculate the impact of REDD projects, since local factors that put a particular area at acute risk of deforestation are a major reason that area gets selected as a project site in the first place.

That rebuttal matters, and it is a fair caution against treating any single investigation as the final word. But the underlying research question did not stay unsettled for long. A peer-reviewed version of the analysis behind the Guardian’s reporting was later published in Science, and it reached a similar conclusion using a formal counterfactual design: the authors examined 26 REDD projects across six countries on three continents and found that most projects had not significantly reduced deforestation.

The lesson for anyone evaluating a project is not that forest offsets are worthless as a category. It is that additionality claims built on a project developer’s own baseline forecast, rather than an independently modeled counterfactual, deserve real skepticism, and that a credit registry approving a methodology is not the same thing as a scientific consensus that the methodology works.

Permanence: The Carbon Was Never Guaranteed to Stay Put

Even a genuinely additional forest credit carries a second, separate problem: the carbon has to stay stored for decades, and forests are increasingly unreliable at that job.

Most nature-based offset programs address this with a buffer pool, a shared reserve of credits set aside to absorb losses if a project burns, gets diseased, or is illegally logged. Each forest offset project contributes a share of its offset credits to the buffer pool, which is meant to compensate for unintentional reversal across all forest projects in the program over a 100-year commitment period. The mechanism is sound in theory. The assumptions behind it are increasingly outdated.

Research from CarbonPlan on California’s forest offset program, the largest of its kind in the United States, found that when fire reversals exhaust more than their fair share of the buffer pool, every other risk category would have to perform better than expected just for the pool to remain solvent, since the entire pool is available to cover any project’s losses regardless of how much that specific project contributed to it.

A separate 2023 study published in PLOS Climate quantified the scale of exposure directly: 26 percent of existing forest carbon offsets in the United States face meaningful wildfire hazard, and improved forest management projects, which represent 96 percent of all credits from forestry projects, span practices with sharply different implications for fire risk, since extending harvest rotations retains higher densities of aboveground biomass that can act as fuel, while thinning reduces fuel loads by removing flammable material.

The most recent research suggests the buffer pools were sized for a climate that no longer exists. A May 2026 study led by University of Utah biologist William Anderegg found that historical conditions suggested around 10 percent of forests would experience wildfire-driven carbon reversals, but under future climate projections that figure climbs to 33 percent nationally, and concluded that wildfire is the largest climate-sensitive risk to the durability of forest-based climate solutions compared with other natural disturbances. Put plainly: the insurance mechanism carbon markets rely on to guarantee permanence was calibrated against a fire regime the planet has already outgrown.

Overcrediting Is Not the Same Problem as Fraud, and That Distinction Matters

One of the more overlooked points in the scientific literature is that most flawed credits are not the product of deliberate fraud. They are the product of methodologies that were approved in good faith, using the best modeling available at the time, and that turned out to be systematically biased once independent researchers checked the results against satellite data.

This distinction has commercial implications that most coverage of the issue skips over. A company that unknowingly bought overcredited offsets is exposed to reputational and, increasingly, legal risk, not because it acted in bad faith but because the underlying verification standard failed. That is precisely the dynamic that has driven a wave of greenwashing litigation across Europe.

Germany’s Federal Court of Justice ruled in June 2024 that “climate neutral” advertising based on offsetting was misleading, in a case brought against confectionery brand Katjes. A separate German court had already banned Lufthansa from advertising a carbon offset scheme that misleadingly suggested its flights could be made carbon neutral.

Regulators have since moved from case-by-case rulings to categorical bans. The EU’s Empowering Consumers for the Green Transition Directive, formally Directive (EU) 2024/825, becomes enforceable on 27 September 2026 and, unlike the German court decisions that preceded it, does not leave room for disclosure-based compromise.

Under the directive, “carbon neutral” or “climate neutral” claims based on offsetting are prohibited in all circumstances, not merely subject to a substantiation test, and companies must remove such claims regardless of the quality of the underlying offsets.

The test instead becomes whether a product’s own value chain, from raw materials through end of life, stores or removes more carbon than it emits through processes within that value chain, which rules out offsetting purchased outside the product’s own supply chain entirely. Notably, the directive does not ban offsetting itself. Companies can still buy carbon credits; they simply cannot translate that purchase into a blanket “carbon neutral” label on a product or brand.

For sustainability and compliance teams, this is the shift that matters more than any single investigative story: the regulatory bar has moved from can you defend this claim if challenged to this claim is automatically unlawful regardless of evidence, and offset quality no longer functions as a legal shield.

What the Research Says Actually Works, and What Does Not

The scientific literature is more nuanced than the headline that offsets “don’t work,” and treating every credit type as equally unreliable is itself a form of misinformation. A few distinctions consistently separate durable value from junk credits.

Removal credits outperform avoidance credits on the additionality question. A ton of carbon dioxide physically pulled out of the atmosphere and stored, through direct air capture, enhanced mineralization, or biochar, does not depend on a counterfactual forecast about what would have happened to a forest. It is measurable after the fact. The tradeoff is cost and scale: actual carbon capture projects that sequester carbon from the atmosphere could work reliably as offsets, but they are currently expensive and operate at small scale.

Avoidance credits, particularly REDD+ forest protection credits, carry the deepest structural additionality problem because they depend on modeling a hypothetical deforestation rate that never happened, a forecast that has repeatedly proven to run hot when checked against satellite evidence.

Contribution claims are gaining ground as a substitute for offset claims. Rather than a company declaring itself carbon neutral because it funded a project elsewhere, a contribution claim states plainly that the company funded climate mitigation without asserting the funding cancels out its own emissions.

The Annual Review of Environment and Resources paper explicitly recommends this shift, arguing the field should focus on creating rules to find and fund the relatively few types of high-quality projects while employing alternative finance and strategies such as contribution claims for critical conservation, renewable energy, and sustainable development projects.

The market itself is already consolidating around this logic. BloombergNEF analysts estimate a smaller but more credible voluntary carbon market could reach 30 billion dollars annually by 2035, while MSCI forecasts up to 35 billion dollars by 2030, with growth concentrated in verified carbon removals and compliance-linked credits rather than the avoidance credits that dominated the market’s earlier growth phase.

A Practical Framework for Evaluating Any Offset Claim

Reduced to a checklist, the research above points toward four questions worth asking of any credit before treating it as a genuine emissions reduction.

Does the baseline come from an independent counterfactual model, or from the project developer’s own forecast? Independent, peer-reviewed baseline modeling has consistently produced lower estimates of avoided deforestation than developer-submitted baselines, in some cases by an order of magnitude.

Is the credit a removal or an avoidance credit, and if it is avoidance, what specifically was at risk? Avoidance credits require proving a negative that never occurred, which is inherently harder to verify than measuring carbon that has actually been captured.

What durability period is the credit rated for, and does the underlying risk model account for a warming climate, or a historical climate baseline that no longer applies? A buffer pool sized against 20th-century wildfire frequency is not a reliable guarantee against 21st-century wildfire frequency.

Is the claim being used to declare neutrality, or to disclose a contribution? A company stating it funded a specific verified project is making a falsifiable, auditable claim. A company declaring itself carbon neutral because of that funding is making a claim regulators in the EU have now decided cannot be substantiated regardless of the offset’s actual quality.

None of this settles the argument over whether carbon markets have a future. It does explain why the scientists most involved in auditing these markets keep raising the same handful of objections, and why the answer from policymakers has increasingly been not to fix the offset claim, but to stop allowing companies to make it.

What People Ask