Why we publish this
An evidence grade nobody can check is an opinion with a letter on it. We publish the formula on the methodology page and the inputs here, so that anyone can recompute our scores from our own data and find out whether we applied our own rules. If your numbers differ from ours, that is a bug report we actually want: editorial@longlifelist.com.
The change log is in the dataset for the same reason. A grade that can move quietly is not a grade, and a change log that is only on a web page is a change log nobody can analyse.
Endpoints
| Endpoint | Returns | Tiers |
|---|---|---|
/v1/interventions | Every published intervention with its aliases, category, kind and current grades per outcome. | Free, Research, Commercial |
/v1/interventions/{key} | One intervention, with every graded outcome, the score breakdown, the four sub-scores and every penalty with its reason. | Free, Research, Commercial |
/v1/grades | The grade table: intervention, outcome, population, grade, evidence score, hype score, formula version, reviewer and review date. | Free, Research, Commercial |
/v1/grades/changes | The grade change log, including each reviewer justification verbatim and the triggering study identifiers. | Free, Research, Commercial |
/v1/studies | Appraised studies: design, species, participants, duration, quality appraisal, preprint and retraction status, DOI and PMID. | Research, Commercial |
/v1/biomarkers | Biomarkers with LOINC codes, UCUM units, panels and published reference ranges with their sources. | Free, Research, Commercial |
/v1/clinics | Directory listings: services, locations, accreditations, verification status and date. No reviews and no ratings. | Commercial |
/v1/costs | Cost guides by therapy and country, with the sample size behind every figure. | Research, Commercial |
/v1/corrections | Our published corrections log, so you can audit our error rate rather than taking our word for it. | Free, Research, Commercial |
All endpoints are read-only, versioned under /v1, paginated with a cursor, and return the formula version alongside every score. Full request and response schemas are in the OpenAPI specification.
Tiers and rate limits
| Tier | Who it is for | Rate limit | Cost |
|---|---|---|---|
| Free | Personal projects, prototypes, journalists checking a figure. Best-effort availability with no commitment. | 60 requests/minute, 10,000/day | Free |
| Research | Academic and non-commercial research. Adds the studies and cost-guide endpoints and bulk dataset exports. | 300 requests/minute, 250,000/day | Free with an academic affiliation and a commitment to cite. We do not charge researchers for evidence appraisal. |
| Commercial | Products and services. Adds the clinic endpoint, webhooks and a published availability target. | Agreed, from 1,000 requests/minute | Priced by call volume |
Limits are per key and returned in response headers, and a 429 tells you when to retry rather than leaving you to guess. If you are about to run a large backfill, email us and we will raise the limit for the window rather than have you discover it at request 10,001.
Licence and citation
CC BY 4.0. Use it commercially, modify it, redistribute it — attribute it.
- Citation format
Use this, replacing the access date:
LongLifeList. LongLifeList Evidence Dataset (evidence formula 1.0.0). 2026. https://longlifelist.com/data. Accessed <date>. Licensed CC BY 4.0.
- When you display a grade
- Show it as a LongLifeList grade, name the outcome it belongs to, include the one-sentence grade meaning, link to the page, and state the date you retrieved it. A bare letter with no outcome and no date is not usable information and is a licence breach.
- When you redistribute the dataset
- Keep the formula version, the review dates and the reviewer attribution on every row. Those fields are what make the row auditable, and stripping them turns published appraisal into an unsourced assertion.
- Models and training
- You may train on the dataset. You may not present its grades as unsourced fact — attribution is the whole point of a graded claim, and a model that reproduces our letters without our citations is producing exactly the kind of confident, unsourced health claim this project exists to counter.
Third-party data we use
The country, state, county and city names on our clinic directory pages, how they nest inside each other, and their populations come from GeoNames, licensed under CC BY 4.0. We re-import it periodically and do not edit the names.
Life expectancy, population aged 65+, physicians per 1,000 and health spending on country pages come from the World Bank World Development Indicators, CC BY 4.0. Each figure shows its year and links to its indicator.
What is deliberately not available
There is no endpoint that returns anything about a person
Not a member, not a patient, not a calculator run, not a review author, not an aggregate derived from health data. This is not a rate-limited or tier-gated capability that a large enough contract unlocks. It does not exist in the API surface at all, and the health database is not reachable from the process that serves the API.
- Reviews and ratings are excluded from the clinic endpoint. A scraped rating detached from its verification status is worse than no rating.
- Draft and unpublished content is excluded. An unreviewed grade is not a grade yet.
- Full study abstracts are excluded where the publisher’s licence does not permit redistribution. You get the identifiers and our appraisal; the abstract is the publisher’s to give.
Is the evidence data really open?
The grades, scores, sub-scores, penalties, study appraisals and the change log are published under CC BY 4.0 with a citation requirement. You can recompute our scores from them, and if your numbers differ we want to hear about it — that is the point of publishing the formula as well as the data.
What is not available?
Anything about a person. There is no endpoint that returns member data, health data, a calculator run or a review author. Reviews and ratings are also excluded from the clinic endpoint, because a scraped rating detached from its verification status is worse than no rating.
Can I use this to build a diagnostic tool?
No. The API terms prohibit presenting a diagnosis, a treatment recommendation or a dose as coming from us, and nothing here is suitable as a dependency in a clinical workflow. The data is appraisal of published research, not patient care.
How stable is the schema?
We give 90 days’ notice before a breaking change and keep the previous version running for that period. Adding a field is not breaking; removing or renaming one is. The formula version travels with every score, so a score that changed because the formula changed is distinguishable from one that changed because the evidence did.
Getting a key
Free keys are self-service from your account settings. For Research access, email hello@longlifelist.com from an institutional address with a sentence on what you are studying. For Commercial, the same address with your expected call volume.
Keep keys server-side. We store only a hash, and a key found in a public client bundle is revoked without notice — see the API terms.