Every quote below was re-fetched directly from bezerocarbon.com for this note (not carried over from an earlier pass). The question this brief asks: not "is BeZero's rating good" (the price-premium research already answers that) but "what does BeZero's own six-factor lens see when it's pointed one layer down, at the funded, proof-verified labor a credit could someday be built from, instead of the aggregated credit itself?"
bezerocarbonmarkets.com, no state-changing call. Captured 2026-07-16; every URL below is independently re-runnable.BeZero's own definitions page states the scale plainly: it represents "the likelihood that a given credit achieves a tonne of CO₂e avoided or removed", nothing else. AAA: "the credit issued has the highest likelihood of achieving 1 tonne of CO₂e avoidance or removal." Each notch down restates the same sentence with one word changed (very high → high → moderate → moderately low → low → very low), down to D: "the lowest likelihood." Co-benefits (biodiversity, social impact) are scored separately and explicitly excluded from this number. That discipline (rate one thing, name what you're not rating) is worth borrowing outright.
Source: bezerocarbon.com/insights/the-bezero-carbon-rating-definition-and-rating-scale (fetched 2026-07-16, exact quotes above)
BeZero's own resources index names its "Risk factor assessment frameworks" explicitly: Additionality risk, Over-crediting risk, Leakage risk, Non-permanence risk, Information risk, and Policy risk, plus a related but distinct Monitoring, Reporting & Verification (MRV) risk-profile category. This is the exact vocabulary Section 7 below applies to a single work packet. (One correction against an earlier internal draft of this note: "perverse incentives" does not appear anywhere in BeZero's own public text: Information risk is the real sixth category, and it turns out to be the most directly relevant one of all.)
Source: bezerocarbon.com/ratings/resources (fetched 2026-07-16)
BeZero's own research states it directly: "each BeZero Carbon ratings notch added an average premium of around 30% to the price a credit commanded in the market," corroborated against an independent dataset (Viridios AI pricing). This is the commercial proof that a rating is doing real work, not decoration: buyers visibly pay for the opinion.
Source: bezerocarbon.com/insights/towards-efficiency-carbon-credit-pricing-and-risk-part-ii (fetched 2026-07-16, exact quote above)
BeZero's project-developer page states its Developer Engagement team practices "open dialogue with project developers before, during and after the rating process" across over 500 projects, backed by 80+ scientists using sector-specific models, closing with unanimous Ratings Committee approval. This is not a self-serve black box: a floor worth naming honestly before proposing anything lighter-weight below.
Source: bezerocarbon.com/ratings/developers (fetched 2026-07-16)
BeZero's free, self-serve Pre-rating Scorecard takes "as little as 20 minutes," asks sector-specific questions across three of the six risk factors (additionality, carbon accounting, permanence), optionally auto-extracts answers from an uploaded project document, and outputs risk-factor indicators plus an overall aaa–d score. Critically, BeZero states outright: "a pre-rating score aligns with the risk drivers outlined in our ratings methodologies, it is not a substitute for a full rating, nor should it be interpreted or communicated in that way." That exact disclaimer (useful signal, explicitly not a certifying rating) is the model the Agent Kit's packet-scoring concept follows.
Source: bezerocarbon.com/insights/introducing-pre-rating-scorecard (fetched 2026-07-16, exact quotes above)
BeZero's public /ratings/listings page shows a preview table ("+680 results"), but the page itself discloses, in its own words: "The projects shown above are illustrative samples created for demonstration purposes only. Project names, IDs, ratings, and all associated data are fictitious and do not represent real projects or published ratings." Real listings sit behind free registration. Worth knowing before anyone (including an earlier draft of this note) cites "+680" as a live rated-project count; it isn't, on the page that's reachable without an account.
Source: bezerocarbon.com/ratings/listings (fetched 2026-07-16)
The packet: tree_planting-…-highland-park-los-angeles-ca-b40-0002, a real, already-defined entry in EcoWealth's ecological work library: not hypothetical, not a superpollutant category still to be authored. Standard: tree_planting_establishment, cited to ANSI A300 Part 6 (Planting) + ISA Best Management Practices for tree planting & establishment + Arbor Day Foundation planting standard; SDG 11.7/15 urban canopy. Wage $95. Proof rule: "GPS waypoint + before/after photos of root flare at grade + mulch/watering-basin confirm + species/staking record." Recurring annually, with a structural sibling work type, tree_establishment_care (see e.g. …-sheridan-wy-b40-0014), that recurs monthly, GPS-matched to the original planting, checking watering/mulch/stake condition and flagging a failing tree to a human steward rather than forcing a claim. Each verdict below is honest about what proof helps with and what it plainly doesn't.
A packet only pays out if postWork (funding) happened before claimWork and submitProofs, tied to this exact GPS point. There's no aggregate counterfactual model to trust: the on-chain timestamp order either shows funding-before-action or it doesn't.
One packet, one wage, one proof: no batch issuance or rounding across a project to inflate. But EWP does not currently convert "one tree planted" into a tCO2e sequestration estimate; that quantification question is a separate, unbuilt layer, honestly out of scope here.
The proof is bound to a specific site (lat/lng, 15km posting radius); a worker can't claim this packet by planting somewhere else. Market-level leakage (would the city have funded this tree anyway, through a different channel) is a project-scale question a single packet can't answer.
A planted tree can still die. EWP's structural answer isn't a claim, it's a recurring proof-gated companion work type: GPS-matched monthly check-ins that record watering/mulch/stake condition and route a failing tree to a human steward instead of quietly keeping the original claim.
The one factor proof discipline was built for. A BeZero rating draws on developer-submitted documents and geospatial estimates, reviewed by analysts. A packet's evidence is a timestamped GPS waypoint, before/after photos, a signature, and an on-chain settlement record, harder to backfill or reshape after the fact than a document review.
The definition itself bakes in point-of-action compliance (an 811 utility-locate stop-work rule, a right-of-way conflict flag to a site steward), but jurisdiction-level policy risk, the kind BeZero's own factor is really about, is a project-scale question a single packet's proof was never built to answer.
Two of six factors plainly don't change at packet scale: non-permanence and policy risk stay real, and over-crediting's quantification question (what's the actual tCO2e a planted tree earns?) is a layer EWP hasn't built yet. But additionality and leakage meaningfully tighten when proof is bound to a funded, GPS-located, timestamped action instead of an aggregated claim, and information risk, the factor an analyst review can never fully escape, is the one a photo, a GPS waypoint, and an immutable settlement record were built to answer directly. None of this claims to replace, compete with, or be rated by BeZero; it's their own lens, credited openly, pointed one level lower in the same stack.
→ See this same packet as a rating card, in BeZero's own palette
