diff --git a/haystack/modeling/evaluation/metrics.py b/haystack/modeling/evaluation/metrics.py
index b56fb473e..9c945827f 100644
--- a/haystack/modeling/evaluation/metrics.py
+++ b/haystack/modeling/evaluation/metrics.py
@@ -30,7 +30,7 @@ except ImportError as exc:
 try:
     from seqeval.metrics import classification_report as token_classification_report
 except ImportError as exc:
-    logger.debug("seqeval could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+    logger.debug("seqeval could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue.")
     token_classification_report = None
 
 
@@ -150,7 +150,9 @@ def compute_report_metrics(head: PredictionHead, preds, labels):
         report_fn = registered_reports[head.ph_output_type]  # type: ignore [index]
     elif head.ph_output_type == "per_token":
         if not token_classification_report:
-            raise ImportError("seqeval could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+            raise ImportError(
+                "seqeval could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue."
+            )
         report_fn = token_classification_report
     elif head.ph_output_type == "per_sequence":
         if not classification_report:
diff --git a/haystack/utils/context_matching.py b/haystack/utils/context_matching.py
index 8be5f9079..2b36f6cd2 100644
--- a/haystack/utils/context_matching.py
+++ b/haystack/utils/context_matching.py
@@ -14,7 +14,7 @@ logger = logging.getLogger(__file__)
 try:
     from rapidfuzz import fuzz
 except ImportError as exc:
-    logger.debug("rapidfuzz could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+    logger.debug("rapidfuzz could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue.")
     fuzz = None  # type: ignore
 
 
@@ -56,7 +56,9 @@ def calculate_context_similarity(
                                  Thus [AB] <-> [BC] (score ~50) gets recalculated with B <-> B (score ~100) scoring ~75 in total.
     """
     if not fuzz:
-        raise ImportError("rapidfuzz could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+        raise ImportError(
+            "rapidfuzz could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue."
+        )
     # we need to handle short contexts/contents (e.g single word)
     # as they produce high scores by matching if the chars of the word are contained in the other one
     # this has to be done after normalizing
diff --git a/haystack/utils/experiment_tracking.py b/haystack/utils/experiment_tracking.py
index 08427d27b..0aee012e9 100644
--- a/haystack/utils/experiment_tracking.py
+++ b/haystack/utils/experiment_tracking.py
@@ -18,7 +18,7 @@ logger = logging.getLogger(__name__)
 try:
     import mlflow
 except ImportError as exc:
-    logger.debug("mlflow could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+    logger.debug("mlflow could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue.")
     mlflow = None
 
 
@@ -165,7 +165,9 @@ class MLflowTrackingHead(BaseTrackingHead):
         Experiment tracking head for MLflow.
         """
         if not mlflow:
-            raise ImportError("mlflow could not be imported. Run 'pip install farm-haystack[eval]' to fix this issue.")
+            raise ImportError(
+                "mlflow could not be imported. Run 'pip install farm-haystack[metrics]' to fix this issue."
+            )
         super().__init__()
         self.tracking_uri = tracking_uri
         self.auto_track_environment = auto_track_environment