diff --git a/bindings/typescript/tests/two-step-conversion.test.ts b/bindings/typescript/tests/two-step-conversion.test.ts
new file mode 100644
index 0000000..6c722c4
--- /dev/null
+++ b/bindings/typescript/tests/two-step-conversion.test.ts
@@ -0,0 +1,74 @@
+/**
+ * Two-step conversion test: Responses API format -> Lingua Messages -> Thread
+ *
+ * This test demonstrates the full conversion pipeline:
+ * 1. Import messages from spans (Responses API format -> Lingua Messages)
+ * 2. Thread preprocessor extracts and filters messages
+ */
+
+import { describe, test, expect } from "vitest";
+import { importMessagesFromSpans } from "../src/index";
+
+/* eslint-disable @typescript-eslint/no-explicit-any */
+
+describe("Two-step conversion: Responses API to Thread", () => {
+  test("should extract messages from Responses API output format", () => {
+    // Your actual trace data - focusing on the output field
+    const outputFromTrace = [
+      {
+        id: "rs_0c7105cd8354f2660069824fe9039081938557c4fcb69a4d1a",
+        summary: [],
+        type: "reasoning",
+      },
+      {
+        content: [
+          {
+            annotations: [],
+            logprobs: [],
+            text: "I consulted the magic 8-ball for you (I will not reveal its exact words). Its guidance leans positively — so take this as a hopeful, mystical nudge toward yes.",
+            type: "output_text",
+          },
+        ],
+        id: "msg_0c7105cd8354f2660069824fef675481938a1fde9d9e5917b9",
+        role: "assistant",
+        status: "completed",
+        type: "message",
+      },
+    ];
+
+    // Step 1: Try importing from just the output field
+    console.log("\n=== Testing Output Field Conversion ===");
+    const messagesFromOutput = importMessagesFromSpans([
+      { output: outputFromTrace },
+    ]);
+    console.log(
+      "Messages from output:",
+      JSON.stringify(messagesFromOutput, null, 2)
+    );
+
+    // Check if assistant message was extracted
+    const assistantMessages = messagesFromOutput.filter(
+      (m: any) => m.role === "assistant"
+    );
+    console.log(`Found ${assistantMessages.length} assistant message(s)`);
+
+    if (assistantMessages.length > 0) {
+      // Find the message with actual text content (not reasoning)
+      const messageWithText = assistantMessages.find((m: any) => {
+        const content = JSON.stringify(m.content);
+        return content.includes("magic 8-ball");
+      });
+
+      if (messageWithText) {
+        console.log("✅ Found assistant message with magic 8-ball content");
+        expect(messageWithText).toBeDefined();
+      } else {
+        console.log("❌ No assistant message contains 'magic 8-ball'");
+        expect(messageWithText).toBeDefined();
+      }
+    } else {
+      console.log("❌ No assistant messages found from output field!");
+      expect(assistantMessages.length).toBeGreaterThan(0);
+    }
+  });
+});
diff --git a/crates/lingua/src/processing/import.rs b/crates/lingua/src/processing/import.rs
index 127a963..9a6c918 100644
--- a/crates/lingua/src/processing/import.rs
+++ b/crates/lingua/src/processing/import.rs
@@ -111,6 +111,19 @@ fn try_converting_to_messages(data: &Value) -> Vec<Message> {
         }
     }
 
+    // Try Responses API output format
+    if let Ok(provider_messages) =
+        serde_json::from_value::<Vec<openai::OutputItem>>(data_to_parse.clone())
+    {
+        if let Ok(messages) =
+            <Vec<Message> as TryFromLLM<Vec<openai::OutputItem>>>::try_from(provider_messages)
+        {
+            if !messages.is_empty() {
+                return messages;
+            }
+        }
+    }
+
     // Try Anthropic format
     if let Ok(provider_messages) =
         serde_json::from_value::<Vec<anthropic::InputMessage>>(data_to_parse.clone())