Carbon Credit Ratings Explained: What a Score Measures and How to Read One

1. What a carbon credit rating is, and what it is not

A carbon credit represents one tonne of CO2 equivalent reduced or removed. Whether the tonne is real depends on the project: whether it would have happened anyway, whether its baseline was inflated, whether the carbon stays stored, and whether the emissions simply moved elsewhere. A carbon credit rating reads the evidence on those questions and turns it into a score.

What a rating is not:

• Not a registry label. Registries and standards issue credits and attach labels, such as the ICVCM's CCP label or a CORSIA eligibility label. A rating is a separate reading of the project by someone else.
• Not a price forecast. A score says how far the evidence supports the credits' claim. It does not say what the credits are worth or will be worth.
• Not a guarantee. A rating reads the documents and data available on a date. New documents, a re-analysis or a change of method can move the score.

2. The risks a rating looks at

There is no universally accepted definition of a high-quality carbon credit (see section 3). The closest shared vocabulary is the Integrity Council for the Voluntary Carbon Market (ICVCM)'s ten Core Carbon Principles, in three groups (ICVCM):

• Governance: effective governance; tracking (a registry that identifies each credit); transparency; robust independent third-party validation and verification.
• Emissions impact: additionality; permanence; robust quantification of emission reductions and removals; no double counting.
• Sustainable impact: sustainable development benefits and safeguards; contribution toward the net zero transition.

Most carbon credit ratings concentrate on the emissions-impact questions, because that is where a credit can fail to represent a tonne:

• Additionality: would the reductions have happened without the money from credits?
• Over-crediting: is the baseline, the «without the project» scenario, set conservatively, or does it inflate the credits?
• Permanence: for stored carbon, such as forests or soils, how likely is a reversal, and is it covered by a buffer?
• Leakage: do the emissions avoided here reappear somewhere else?
• Double counting: is the same tonne issued, claimed or used twice?

Co-benefits and safeguards (effects on communities and nature) are assessed by some providers and not by others.

3. Why two providers can rate the same project differently

Each provider publishes its own framework, with its own scale, its own tests and its own weights, so their scores are not directly comparable.

A September 2023 report by Perspectives Climate Group, funded by Carbon Market Watch, compared the methods of four carbon credit rating providers: BeZero, Calyx, Renoster and Sylvera (Carbon Market Watch, 2023). It found that:

• «The lack of a universally accepted and in-depth definition of carbon credit quality is a fundamental issue for the carbon market» (p. 1);
• «Approaches to rate carbon credits differ in many ways», including how a weak additionality result limits the overall score, and how double issuance and double counting, co-benefits and safeguards, leakage, buffer strength and permanence are assessed (p. 1);
• where the same projects were rated by several providers, the results could diverge widely: one REDD+ project was placed in the top tier by one provider and near the bottom of the scale by two others, using ratings retrieved on 2 June 2023 (pp. 56–58).

For a buyer, this means a score is only as useful as the method behind it. Before relying on any score, check what it measures, what evidence it read, and how recent that evidence is.

4. How CarbonMeld scores a project

CarbonMeld's score is CarbonMeld's own document-based assessment, for information only: it is not investment advice, and it is not a rating from the registry or from the standard. The method is published in full on the methodology page:

1. Documents. The registry's records and the project's public documents: design, validation, monitoring and verification reports.
2. Facts. Values are read one at a time. Each value is kept only when the passage it comes from is found in the source pages.
3. Scores. Four dimensions, each scored from 0 to 10 against the project's own methodology:
   • Integrity, 35%: additionality, the quality and conservatism of the baseline, permanence risk, leakage.
   • Transparency, 25%: public data, accessible monitoring, a public design document, a findable registry record.
   • Claim safety, 25%: greenwashing risk for the buyer, ICVCM CCP eligibility, the age of the credits' vintage.
   • Documentation, 15%: whether design documents, monitoring reports and third-party verification reports are available.
4. Gates. Sector deductions, then caps, then the overall score: integrity × 0.35 + transparency × 0.25 + claim safety × 0.25 + documentation × 0.15.
5. Published page. Each new text is checked against the verified facts before it is published.

Two further values say how much the score can be relied on. Both are computed by code from what was read, and neither changes the score itself:

• Evidence scope: whether registry documents were read, only the registry's own records, or only a marketplace listing.
• Confidence: high, medium or low, from the project's stage and the documents read. As the method page puts it, «a 6.8 with high confidence is not equivalent to a 6.8 with low confidence».

5. Reading one CarbonMeld page, step by step

The example is the Haicheng Landfill Gas Power Generation Project in China, Verra registry ID VCS-2455. Its verification report is the one read page by page in our guide How to read a verification report. Everything below is what any visitor sees on the page, as read on 7 November 2026.

1. The overall score: 7.7/10. It sits under the heading «CarbonMeld document-based assessment»: CarbonMeld's own assessment, for information only, not investment advice and not a rating from Verra or from the standard.
2. The confidence: medium. Read it together with the score. «High» needs the project to be at a verified stage, at least one monitoring or verification report and at least two of its documents read, and no rejected answer of the analysis model counting against it. Stale monitoring data lowers confidence by one level. The methodology page gives the full rule.
3. The version and the date: «Methodology v2.2 · analysed on 16 Oct 2026». A score belongs to one version of the method. When a project waits for re-analysis, its page says which earlier version its score comes from.
4. The evidence: «24/24 registry documents read». The second number is how many of the project's registry documents CarbonMeld holds, and the first is how many of them the analysis read. A gap between them is worth noticing.
5. The registry labels, with their source and date. The page shows what the registry says, read on 1 Nov 2026: Haicheng's credits carry the ICVCM CCP label on the registry, and the project has no CORSIA label and no Article 6 authorization. These are the registry's labels, not CarbonMeld's.
6. The methodology versions. For each monitoring period the page shows the version of ACM0001 that the documents state (v19.0), and that it falls within the version range the ICVCM approved for VCS.
7. What is Premium. The four dimension scores, the gates applied to this score and the reasons behind any red flags are Premium. The overall score, confidence, version, date and evidence count are free for every project with a current assessment.

To compare projects, start from carbon credit ratings by project, or from the boards that rank projects of one kind, such as landfill gas and other methane projects or Verra (VCS) projects.

6. Limits of any score

• It reads documents, not the site. CarbonMeld does not visit projects. If the documents are missing, old or wrong, the score can only reflect that.
• Documents can be missing or late. Registries publish with delays, and some projects publish little. On a page with a current analysis, CarbonMeld shows how many of the project's documents it holds and how many of them were read, and it scores a dimension with insufficient information between 4 and 5, not at the top.
• Scores change. A new monitoring report, a re-analysis on a newer method version or a registry status change can move a score up or down.
• It is information, not advice. Before a purchase or a public claim, check the registry record itself, the labels the credits actually carry, and the rules of the claim you want to make. Our due-diligence checklist goes document by document.

• Carbon credit ratings by project
• Our methodology
• Landfill gas and other methane projects, ranked
• Verra (VCS) projects, ranked
• How to read a verification report
• Carbon credit due diligence: a registry document checklist