Carbon offsets promise that a tonne of emissions here can be balanced by a tonne cut or removed somewhere else. Whether that works depends on whether the tonne was real, would not have happened anyway, and stays out of the atmosphere. Over the past few years, independent researchers have tested those claims against real-world data. This page summarises what they found, what the criticism of that research says, and how the market has responded.
It is part of our Quality section.
What does the research find?
Three peer-reviewed studies are among the most cited.
Forest protection (REDD+). West and colleagues examined 26 avoided-deforestation project sites in six countries, using synthetic control methods to estimate what would have happened without them. They found that most projects had not significantly reduced deforestation, and that where they had, the reductions were substantially lower than claimedSource 1. The gap came from the difference between the baselines projects set in advance and what actually happened in comparable control areas, and the authors called for urgent revisions to how deforestation baselines are builtSource 1.
Cookstoves. Gill-Wiehl, Kammen and Haya compared five cookstove methodologies with published evidence and their own analysis. They found the methodologies out of line with the evidence on several factors, in order of importance: the fraction of non-renewable biomass, firewood-to-charcoal conversion, stove adoption, stove usage, fuel consumption, “stacking” (households using more than one stove), rebound and emission factorsSource 2. They estimated that their sample of projects was over-credited 9.2 times. Gold Standard’s metered methodology, which directly monitors fuel use, was the most closely aligned with their estimates, at 1.5 times over-creditedSource 2. The authors note that additionality, leakage, permanence and overlapping claims need more research, so their estimate covers only part of the pictureSource 2.
Across project types. Probst and colleagues synthesised 14 studies covering 2,346 projects, about a fifth of all credits issued to date, almost 1 billion tonnes of CO₂e. They also drew on 51 studies of similar interventions carried out without carbon credits. They estimated that less than 16% of the credits issued to those projects were real emission reductionsSource 3.
| Project type | Real reductions (% of credits) |
|---|---|
| HFC-23 abatement | 68 |
| Avoided deforestation | 25 |
| SF6 destruction | 16 |
| Cookstoves | 11 |
| Wind power | Not significant |
| Improved forest management | Not significant |
As of Nov 2024 · Source: [3] Probst et al., Nature Communications 2024
The authors concluded that carbon crediting mechanisms need to be reformed fundamentally to contribute meaningfully to climate mitigationSource 3.
The other side of the debate
These findings have been challenged. In January 2023, Verra published a technical review of an earlier version of the West et al. analysis and other studies. It called them “patently unreliable”, arguing that they ignored socio-economic drivers of deforestation, compared projects with areas under little threat, used too few comparison areas and relied on unsuitable deforestation dataSource 4.
Each study looks at particular project types, methodologies and time periods, and some of the methodologies involved have since been replaced or refused the CCP label (see below). Method choices matter a great deal in this research, and the debate over how to set baselines continues. Baselines decide how many credits a project earns, which is why the ICVCM’s approval of new forest methodologies focused on them (see below)Source 6.
How the market is responding
The main response has been to tighten the rules for counting credits.
- A common quality bar. The ICVCM’s 10 Core Carbon Principles, covering areas such as additionality, permanence and robust quantification, set a threshold that methodologies must meet to carry the CCP labelSource 5.
- Forest protection. In November 2024 the ICVCM approved three REDD+ methodologies. Under Verra’s new VM0048, developers no longer set their own baselines from self-selected reference areas; credits from Verra’s older REDD+ methodologies won’t be able to receive the CCP labelSource 6.
- Cookstoves. In March 2025 it approved three cookstove methodologies, including Gold Standard’s metered methodology, with conditions on how fuel use is measured, and found two earlier methodologies insufficiently rigorousSource 7.
- Renewable energy. In August 2024 it decided that credits from eight current renewable energy methodologies would not receive the label, because they don’t test rigorously enough whether projects would have gone ahead without carbon financeSource 8.
- How credits are used. VCMI says credits must be used in addition to, and not to delay or displace, deep cuts in a company’s own emissionsSource 9.
What this means if you buy credits
The evidence doesn’t mean every credit is worthless. It means the label “offset” tells you little on its own, and what you buy matters.
- Cut first. The Oxford Offsetting Principles start with reducing your own emissions and treat credits as a complement for what remainsSource 10.
- Prefer stronger methods. Look for credits from methodologies approved against the Core Carbon Principles, and check how the project measures its results.
- Favour durable removals over time. The Oxford Principles recommend shifting toward removals with a low risk of reversalSource 10. See removal vs avoidance credits.
- Verify what you buy. Check retirement and serial numbers yourself: our guide to how to verify carbon credits shows how, and carbon credit scams lists the warning signs.
Claim what you can prove
Given the evidence, avoid claims that a purchase cancels your emissions. A factual statement, such as the number of tonnes you funded, the project and a link to the retirement record, is more honest and easier to defend.