diff --git a/app/services/jsonld.py b/app/services/jsonld.py
index 216e682..22dccdd 100644
--- a/app/services/jsonld.py
+++ b/app/services/jsonld.py
@@ -613,6 +613,11 @@ def map_dataset_to_dcat(
         "keywords": dataset.keywords.split(",") if dataset.keywords else [],
         "license": dataset.license,
         "version": dataset.version,
+        # The device and shot the dataset is filed under, as plain names rather
+        # than links, so that other catalogues can filter on them whatever
+        # addresses they give devices and shots.
+        "fuel:device": dataset.device_name,
+        "fuel:shot": dataset.shot_id,
     }
 
     if dataset.persistent_identifier:
diff --git a/docs/dcat-jsonld.md b/docs/dcat-jsonld.md
index a674e27..b9b58f2 100644
--- a/docs/dcat-jsonld.md
+++ b/docs/dcat-jsonld.md
@@ -144,6 +144,26 @@ The record has no route of its own: its identifier is the resource's with `#reco
 resolves to the document that holds it. A catalogue that lists other resources, such as the
 catalogue of devices, embeds their nodes without their records.
 
+## Device and shot
+
+A dataset filed under a device, or under one of the device's shots, names them in `fuel:device`
+and `fuel:shot`:
+
+```json
+{
+  "@type": "dcat:Dataset",
+  "@id": "https://example.org/datasets/13",
+  "title": "Equilibrium (Shot 30421)",
+  "fuel:device": "mast",
+  "fuel:shot": "30421"
+}
+```
+
+The values are the device's `name` and the shot's ID as plain text, not links to the device and
+shot documents. Other catalogues give the same device and shot their own addresses, but can use
+the same names, so a search across catalogues can filter on these values. A dataset attached to a
+device but no shot has `fuel:device` only, and a global dataset has neither.
+
 ## Dataset vs Distribution
 
 FDS follows the [W3C DCAT ontology](https://www.w3.org/TR/vocab-dcat/): a **`dcat:Dataset`** is the abstract metadata entity describing *what* the data is, while a **`dcat:Distribution`** is a concrete physical access path describing *how* to retrieve it.
@@ -393,7 +413,7 @@ See [Provenance](provenance.md) for the model behind these terms.
 | `oa` | `http://www.w3.org/ns/oa#` | Web Annotation, used by `dqv:QualityAnnotation` (`oa:motivatedBy`, `oa:hasBody`) |
 | `adms` | `http://www.w3.org/ns/adms#` | `adms:identifier`, a persistent identifier registered elsewhere |
 | `skos` | `http://www.w3.org/2004/02/skos/core#` | `skos:notation`, the value of that identifier |
-| `fuel` | `https://w3id.org/fuel/ns#` | Fusion Energy Lexicon, the role concepts used by `prov:hadRole` and `dcat:hadRole` |
+| `fuel` | `https://w3id.org/fuel/ns#` | Fusion Energy Lexicon: the role concepts used by `prov:hadRole` and `dcat:hadRole`, and a dataset's `fuel:device` and `fuel:shot` |
 
 ## Ontology mapping summary
 
@@ -412,6 +432,8 @@ See [Provenance](provenance.md) for the model behind these terms.
 | "dataset produced by" | `prov:wasGeneratedBy` | PROV-O |
 | "activity used input / instrument" | `prov:used` + `prov:qualifiedUsage` (`prov:hadRole`) | PROV-O |
 | "activity associated with agent" | `prov:wasAssociatedWith` + `prov:qualifiedAssociation` (`prov:hadRole`) | PROV-O |
+| Dataset `device_name` | `fuel:device`, as plain text | FuEL |
+| Dataset `shot_id` | `fuel:shot`, as plain text | FuEL |
 | `persistent_identifier` | `adms:identifier` → `adms:Identifier` | ADMS, DCAT-AP |
 | `issued` (publication date) | `dct:issued` on the resource | Dublin Core, DCAT 3 |
 | producing Activity's `ended_at` | `prov:generatedAtTime` on the resource | PROV-O |
diff --git a/tests/api/v1/test_dataset_router.py b/tests/api/v1/test_dataset_router.py
index bad98c6..b0e5a3f 100644
--- a/tests/api/v1/test_dataset_router.py
+++ b/tests/api/v1/test_dataset_router.py
@@ -913,3 +913,51 @@ def test_temporal_coverage_roundtrip(
     assert cov["@type"] == "dct:PeriodOfTime"
     assert cov["startDate"].startswith("2024-03-15T14:00:00")
     assert cov["endDate"].startswith("2024-03-15T14:30:00")
+
+
+def test_dataset_jsonld_names_its_device_and_shot(
+    test_client: TestClient, session: Session, admin_user_token: dict
+):
+    DeviceService(session).create(
+        DeviceCreate(name="MAST", type="Spherical"), user=admin_user
+    )
+    ShotService(session).create(
+        ShotCreate(id="30421", device_name="MAST"), user=admin_user
+    )
+    datasets = DatasetService(session)
+    on_shot = datasets.create(
+        DatasetCreate(
+            name="efit",
+            device_name="MAST",
+            shot_id="30421",
+            url="s3://mast/30421/efit",
+        ),
+        user=admin_user,
+    )
+    on_device = datasets.create(
+        DatasetCreate(name="geometry", device_name="MAST", url="s3://mast/geometry"),
+        user=admin_user,
+    )
+    on_nothing = datasets.create(
+        DatasetCreate(name="constants", url="s3://global/constants"),
+        user=admin_user,
+    )
+    session.commit()
+    ld = {**admin_user_token, "Accept": "application/ld+json"}
+
+    shot_node = resource(
+        test_client.get(f"/v1/datasets/id/{on_shot.id}", headers=ld).json()
+    )
+    device_node = resource(
+        test_client.get(f"/v1/datasets/id/{on_device.id}", headers=ld).json()
+    )
+    global_node = resource(
+        test_client.get(f"/v1/datasets/id/{on_nothing.id}", headers=ld).json()
+    )
+
+    assert shot_node["fuel:device"] == "mast"
+    assert shot_node["fuel:shot"] == "30421"
+    assert device_node["fuel:device"] == "mast"
+    assert "fuel:shot" not in device_node
+    assert "fuel:device" not in global_node
+    assert "fuel:shot" not in global_node