SDGs
The Sustainable Development Goals (SDGs) are the 17 global goals the United Nations adopted in 2015 to address challenges like poverty, inequality, climate change and environmental degradation, from No poverty (SDG 1) to Partnerships for the goals (SDG 17). OpenAlex tags each work with the goals it addresses, so you can find and analyze research on a given global challenge: for example, every work tagged with SDG 13 (Climate action), to study climate-research trends. There are exactly 17; SDGs are one of several aboutness signals OpenAlex offers. An SDG’s OpenAlex ID looks like https://openalex.org/sdgs/3; list them all at api.openalex.org/sdgs.
About
The 17 goals themselves are a fixed UN vocabulary, so this section is about the tagging: how OpenAlex decides which works address which goal.
Since October 2026, SDG tags come from a classifier OpenAlex trained and tested itself. A large language model read the titles and abstracts of 200,000 works and decided, goal by goal, whether each work addresses that goal. We trained a small classifier on those decisions and run it on every work’s vector, a numeric summary of its title and abstract. For each goal it gives a score from 0 to 1. A work’s sustainable_development_goals list holds every goal scoring 0.4 or above, and those lists add up to each SDG’s works_count and cited_by_count. It replaced the Aurora SDG classifier, whose tags stay in a deprecated field until November 2026.
How well it works
The new tags are right far more often than the old ones. We drew 2,000 works at random from OpenAlex and had a committee of AI judges decide every goal for every work against the UN’s goal and target text: two judges independently, and a third where they disagreed. Against those decisions the new tags score F1 0.70 (precision 0.68, recall 0.72); Aurora’s scored 0.29.
On a test with no model in the loop, the survey in which researchers voted on whether papers contribute to a goal (8,767 paper-and-goal questions with a clear majority, collected by the Aurora project), the new classifier scores F1 75.5 to Aurora’s 63.6.
Works with a title but no abstract are tagged too, a little less accurately: about 65% of their tags are right.
What score means
score is the classifier’s confidence, not a probability. It runs from 0 to 1, a goal appears on a work at 0.4 or above, and a higher score means the classifier is surer. When the judges checked, tags scored 0.4 to 0.5 were right about a quarter of the time, tags scored 0.5 to 0.8 about half the time, 0.8 to 0.9 about 7 times in 10, and above 0.9 about 8 times in 10. Use it to rank works or to keep only the surest tags in your own analysis.
What counts as addressing a goal
A work counts when it addresses one of the goal’s UN targets, not when it only touches the goal’s theme. SDG 15 (Life on land), for example, is about protecting and restoring land ecosystems, forests, soils and biodiversity: a study of deforestation, land degradation or invasive species is SDG 15, but a species occurrence record, a specimen record or a species description with no conservation or ecosystem aim is not. Mentioning a theme in passing is not addressing it.
Things to keep in mind
- It reads the title and abstract only, and it can be wrong. Treat SDG tags as a broad, comparable filter, not a verdict on a single paper.
- Counts per goal changed at the switch, some of them a lot. If you compare SDG numbers from before October 2026 with numbers from after, say which classifier each came from.
updated_datedid not change for the switch. The new tags reached most works without changing theirupdated_date. If you keep a copy current withfrom_updated_dateor by downloading only changed snapshot partitions, the new tags will not reach you that way: reloadsustainable_development_goalsfor all works once, from a full snapshot released after the switch or from the API.
The old Aurora tags (deprecated)
Aurora’s tags stay available until November 2026, in their own field. Every work carries sustainable_development_goals_aurora: the Aurora tags as they stood in October 2026, in the same shape (id, display_name, score). It is output only: you can read it and select it, but you cannot filter, sort or group by it. It is deprecated and will be removed in November 2026, so use sustainable_development_goals, and save anything you need from the old tags before then.
Questions, or a tag that looks wrong: [email protected].
Attributes
This is the canonical dictionary of every attribute on an SDG object. Attributes shared with other entities are documented once on Common attributes and linked below.
id
String. The OpenAlex ID for this goal, e.g. https://openalex.org/sdgs/3. Unlike most entities, the numeric part is just the goal number (1 to 17). See Common attributes.
display_name
String. The goal’s name, e.g. Good health and well-being. See Common attributes.
description
String. A one-sentence statement of the goal, e.g. “Ensure healthy lives and promote well-being for all at all ages.”
ids
Object. External identifiers for this goal as URIs. SDG-specific keys: openalex, un (the goal’s UN metadata URI, e.g. https://metadata.un.org/sdg/3), and wikidata.
image_url
String. URL of the goal’s official UN icon (an SVG on Wikimedia Commons).
image_thumbnail_url
String. The same icon as image_url, scaled down (width=300).
works_count
Integer. How many works are tagged with this goal (scored 0.4 or above for it). See Common attributes.
cited_by_count
Integer. Total citations across all works tagged with this goal. See Common attributes.
works_api_url
String. A ready-made Works API URL that returns every work tagged with this goal: works?filter=sustainable_development_goals.id:<un-uri>.
created_date
String. The date this goal was added to OpenAlex (YYYY-MM-DD). See Common attributes.
updated_date
String. The ISO 8601 UTC timestamp of the last change to this object. See Common attributes.
Of these, works_count, cited_by_count, display_name, and id can be used as filters and sort keys; works_count, cited_by_count, and id also support group_by. The others are select-only columns.
In the API
The SDGs endpoint is at api.openalex.org/sdgs, a short, fixed list of 17. Fetch a single goal by number (/sdgs/3) or the whole list, and filter, sort, and group over the fields above. For the full list of endpoints see the endpoints index.
The more common way to use SDGs is from the works side: every work carries a sustainable_development_goals list, and you can filter works by goal.
# List all 17 SDGs
https://api.openalex.org/sdgs
# One goal
https://api.openalex.org/sdgs/13
# Works tagged with SDG 3 (Good health and well-being)
https://api.openalex.org/works?filter=sustainable_development_goals.id:3Last updated
View as Markdown