Series: Repricing Carbon — A Buyer’s Guide to the New Voluntary Market
- The Year Carbon Credits Stopped Being Free
- Inside a Carbon Credit: What “High Quality” Actually Means (you are here)
- Ratings Aren’t New. Carbon Ratings Are.
- How a Carbon Buyer Thinks in 2026
- The Reform Loop
The carbon market spent most of 2024 talking about quality. Almost none of that conversation defined the word.
“High-integrity credits.” “High-quality offsets.” “Defensible carbon.” The vocabulary moved faster than the standards behind it. A buyer reading procurement documents in early 2026 will see the same adjectives repeated across three different vendors offering three different things.
This piece is the definition. There are five questions that determine whether a carbon credit is what it claims to be. Most credits in circulation fail at least one. A few fail four. The buyer’s job is to know which question is being failed before signing the purchase order.
Question one: would this have happened anyway?
This is additionality. It is the question that does the most work in any honest credit evaluation, and it is also the easiest to wave away.
A carbon credit is supposed to represent emissions that did not happen because the project happened. The project is supposed to be the cause. If the avoided emissions would have been avoided without the credit, the credit is paying for something that was going to occur regardless. The atmosphere is no different. The buyer is paying for nothing.
Additionality breaks in predictable ways. A hydropower project in a country that builds hydropower as a matter of national policy is probably not additional. A forest preservation project on land that was not under credible threat is probably not additional. A renewable energy project in a market where renewables are already the cheapest power source is almost certainly not additional.
The methodology may still approve the project. The registry may still issue the credits. The credits will exist, will be tradeable, and will be retired by a buyer who believed they were buying climate impact. They will represent close to zero.
This is the failure mode that drove the Verra story in 2023. The REDD+ methodology allowed projects to claim a baseline rate of deforestation that, in many cases, never would have occurred. The credits were not fraudulent. They were non-additional. The distinction does not matter to the buyer’s general counsel.
Question two: what is being compared to what?
This is the baseline question. It is the technical core of every quality dispute in the market.
A carbon credit is a comparison. The project’s emissions outcome is compared to a counterfactual, what would have happened in the absence of the project. The credit measures the gap. The baseline is the counterfactual.
The baseline is also, almost always, a model. No one can observe the world that did not happen. So the baseline is built from historical data, regional benchmarks, deforestation rates, fuel mixes, business-as-usual scenarios. Each of those inputs is a judgment. Each judgment can be aggressive or conservative.
An aggressive baseline assumes the worst-case counterfactual. The forest would have been clear-cut. The grid would have run on coal. The cookstove would have burned wood at three meals a day. The project then looks heroic by comparison, and generates more credits.
A conservative baseline assumes a realistic counterfactual. The forest faced moderate pressure. The grid was already shifting. The cookstove was used at one meal. The project looks smaller, generates fewer credits, but the credits it generates are credible.
The same project can produce dramatically different credit volumes depending on which baseline a developer chooses, and which the verifier accepts. The methodology sets the rules for choosing. The rules have, historically, allowed a lot of room.
When you hear that a project “over-credited” or “inflated its baseline,” this is the mechanic. No one is forging numbers. The numbers are real, in the sense that they were calculated from a defensible model. The model just happened to be the most generous defensible model available.
Question three: is the carbon staying out of the atmosphere?
This is permanence. It is the question nature-based projects struggle with the most.
A ton of CO2 emitted by a power plant is gone. There is no version of the future where it un-emits itself. A ton of CO2 stored in a tree is conditionally absent. The tree can burn. The forest can be cut. The soil can release what it sequestered. The carbon that the credit represented can return to the atmosphere, and often does.
The market has a few answers to this. Buffer pools, where a percentage of issued credits is held back to cover reversal risk across the portfolio. Long-term monitoring obligations. Insurance products. Permanence claims expressed in years (100 years, 1,000 years, “geological”).
None of them eliminate the asymmetry. A forest credit and a direct air capture credit, both nominally one ton, are not equivalent assets. They have different return profiles. They carry different risks. The market is starting to price this distinction. Most credit accounting still does not.
This is why removal credits trade at a premium to avoidance credits, even at the same nominal carbon volume. The buyer is not paying more for ideology. The buyer is paying more for a tighter permanence guarantee.
Question four: did the emissions just move?
This is leakage. It is the question regulators are starting to ask the loudest.
A protected forest in one place does not help the climate if the demand for timber moves to a forest two valleys over. A reduced-emissions factory does not help the climate if the displaced production moves to a less-regulated factory in another country. A fuel-switch project does not help the climate if the displaced fuel is sold to a buyer who would not otherwise have used it.
Carbon does not respect project boundaries. Leakage is the difference between what the project did and what the global emissions ledger actually changed.
Good methodologies require leakage adjustments. The project’s credit volume is reduced by an estimated leakage factor. The factor is, again, a model. The model can be conservative or generous. Even when it is conservative, it tends to underestimate. Leakage is hard to measure because the displacement happens somewhere the project is not looking.
Engineered removal projects largely avoid this question. A direct air capture facility does not displace anyone. A biochar plant does not relocate anyone’s emissions. This is one of the structural reasons removals are easier to defend than avoidance.
Question five: can the numbers be trusted?
This is MRV, measurement, reporting, verification. It is the procedural backbone of a credit. It is also the part of a credit a buyer rarely reads in detail.
Every credit is generated against a measurement protocol. The protocol specifies what gets measured, how often, by whom, with what equipment, against what calibration standard. The protocol also specifies how the measurements are reported, who reviews them, and how often verification visits occur.
MRV quality varies enormously between project types and between projects within a type. A renewable energy project on a grid-connected meter has straightforward MRV. The meter reads kilowatt hours. The emissions factor for the grid is published. The math is mechanical.
A soil carbon project has the opposite situation. Soil sampling is expensive, sparse, and noisy. Modeling fills most of the gap. The MRV is doing real scientific work, with real uncertainty bands, and the credit is being treated by the market as if the uncertainty does not exist.
A buyer who does not look at MRV is taking the methodology on faith. The methodology is taking the verifier on faith. The verifier is taking the project developer’s data on faith. The chain is only as strong as the audit at the bottom.
Why a credit can pass the methodology and still be a bad credit
This is the part that surprises buyers who are new to the market.
A credit can be issued under a valid methodology, by a credited project developer, with full registry approval, and still fail two or three of the five questions above. The methodology was followed. The execution was poor. The credit exists, is legal, and is for sale.
The registry will not catch this. The registry’s job ends at “did the project follow the methodology.” The market’s job is to catch what the registry does not.
This is the gap that buyer-side ratings firms exist to fill. They evaluate the project as a project, not the methodology as a methodology. They ask whether the baseline is realistic, whether the additionality argument holds up, whether the permanence claim is defensible, whether the leakage estimate is honest, whether the MRV is sound. They produce a rating that is separable from the registry’s stamp.
Not all ratings firms ask all five questions equally well. The market is not yet uniform. Methodologies differ, project types differ, and the depth of evaluation a rating represents differs from firm to firm.
That is what Part 3 is about. Ratings are not a single, settled product. They are a category that is being built in real time, the way bond ratings were built a hundred years ago.
The question is not whether to use ratings. The question is which rating, by whom, with what methodology, and reviewed against what evidence.
A note from the author. I am a writer who cares about sustainability, and when it comes to carbon credits I am still very much a learner. There are a lot of people who know this market far better than I do, and I have real respect for the work they have put into building it. If I got something wrong in here, I apologize, and I would genuinely like to hear about it so I can learn and correct it. I am writing this to start a conversation, not to have the last word. That is the whole point. This is a learning experience for me too, and the conversation is what moves all of us forward. If this piece helped you, share it. If you see it differently, even better. Let’s talk.
