diff --git a/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/ActivityExtensions.cs b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/ActivityExtensions.cs
new file mode 100644
index 0000000..8f984af
--- /dev/null
+++ b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/ActivityExtensions.cs
@@ -0,0 +1,167 @@
+using System.Diagnostics;
+
+namespace ANcpSdk.AspNetCore.ServiceDefaults.Instrumentation.GenAi;
+
+/// <summary>
+///     Extension methods for setting GenAI semantic convention attributes on Activities.
+/// </summary>
+/// <remarks>
+///     Provides fluent API for OTel 1.39 GenAI semantic conventions.
+/// </remarks>
+public static class GenAiActivityExtensions
+{
+    /// <summary>
+    ///     Sets GenAI request attributes on the activity.
+    /// </summary>
+    /// <param name="activity">The activity to set tags on.</param>
+    /// <param name="model">The model requested (e.g., "gpt-4o", "claude-3-opus").</param>
+    /// <param name="temperature">Temperature setting (0.0-2.0).</param>
+    /// <param name="maxTokens">Maximum tokens to generate.</param>
+    /// <param name="topP">Nucleus sampling threshold.</param>
+    /// <param name="topK">Top-k sampling parameter.</param>
+    /// <returns>The activity for fluent chaining.</returns>
+    public static Activity SetGenAiRequest(
+        this Activity activity,
+        string? model = null,
+        double? temperature = null,
+        int? maxTokens = null,
+        double? topP = null,
+        int? topK = null)
+    {
+        ArgumentNullException.ThrowIfNull(activity);
+
+        if (model is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.RequestModel, model);
+
+        if (temperature.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.RequestTemperature, temperature.Value);
+
+        if (maxTokens.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.RequestMaxTokens, maxTokens.Value);
+
+        if (topP.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.RequestTopP, topP.Value);
+
+        if (topK.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.RequestTopK, topK.Value);
+
+        return activity;
+    }
+
+    /// <summary>
+    ///     Sets GenAI token usage attributes on the activity.
+    /// </summary>
+    /// <param name="activity">The activity to set tags on.</param>
+    /// <param name="inputTokens">Number of input/prompt tokens.</param>
+    /// <param name="outputTokens">Number of output/completion tokens.</param>
+    /// <param name="cachedTokens">Number of cached input tokens (Anthropic).</param>
+    /// <param name="reasoningTokens">Number of reasoning tokens (o1-style models).</param>
+    /// <returns>The activity for fluent chaining.</returns>
+    public static Activity SetGenAiUsage(
+        this Activity activity,
+        long? inputTokens = null,
+        long? outputTokens = null,
+        long? cachedTokens = null,
+        long? reasoningTokens = null)
+    {
+        ArgumentNullException.ThrowIfNull(activity);
+
+        if (inputTokens.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.UsageInputTokens, inputTokens.Value);
+
+        if (outputTokens.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.UsageOutputTokens, outputTokens.Value);
+
+        if (cachedTokens.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.UsageInputTokensCached, cachedTokens.Value);
+
+        if (reasoningTokens.HasValue)
+            activity.SetTag(SemanticConventions.GenAi.UsageOutputTokensReasoning, reasoningTokens.Value);
+
+        return activity;
+    }
+
+    /// <summary>
+    ///     Sets GenAI response attributes on the activity.
+    /// </summary>
+    /// <param name="activity">The activity to set tags on.</param>
+    /// <param name="model">The model that generated the response.</param>
+    /// <param name="responseId">Unique completion identifier.</param>
+    /// <param name="finishReasons">Reasons the model stopped generating.</param>
+    /// <returns>The activity for fluent chaining.</returns>
+    public static Activity SetGenAiResponse(
+        this Activity activity,
+        string? model = null,
+        string? responseId = null,
+        string[]? finishReasons = null)
+    {
+        ArgumentNullException.ThrowIfNull(activity);
+
+        if (model is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ResponseModel, model);
+
+        if (responseId is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ResponseId, responseId);
+
+        if (finishReasons is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ResponseFinishReasons, finishReasons);
+
+        return activity;
+    }
+
+    /// <summary>
+    ///     Sets GenAI agent attributes on the activity (OTel 1.38+).
+    /// </summary>
+    /// <param name="activity">The activity to set tags on.</param>
+    /// <param name="agentId">Unique agent identifier.</param>
+    /// <param name="agentName">Human-readable agent name.</param>
+    /// <param name="agentDescription">Agent description.</param>
+    /// <returns>The activity for fluent chaining.</returns>
+    public static Activity SetGenAiAgent(
+        this Activity activity,
+        string? agentId = null,
+        string? agentName = null,
+        string? agentDescription = null)
+    {
+        ArgumentNullException.ThrowIfNull(activity);
+
+        if (agentId is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.AgentId, agentId);
+
+        if (agentName is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.AgentName, agentName);
+
+        if (agentDescription is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.AgentDescription, agentDescription);
+
+        return activity;
+    }
+
+    /// <summary>
+    ///     Sets GenAI tool attributes on the activity (OTel 1.39).
+    /// </summary>
+    /// <param name="activity">The activity to set tags on.</param>
+    /// <param name="toolName">Tool name.</param>
+    /// <param name="toolCallId">Tool call identifier.</param>
+    /// <param name="conversationId">Session/thread identifier for multi-turn conversations.</param>
+    /// <returns>The activity for fluent chaining.</returns>
+    public static Activity SetGenAiTool(
+        this Activity activity,
+        string? toolName = null,
+        string? toolCallId = null,
+        string? conversationId = null)
+    {
+        ArgumentNullException.ThrowIfNull(activity);
+
+        if (toolName is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ToolName, toolName);
+
+        if (toolCallId is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ToolCallId, toolCallId);
+
+        if (conversationId is { Length: > 0 })
+            activity.SetTag(SemanticConventions.GenAi.ConversationId, conversationId);
+
+        return activity;
+    }
+}
diff --git a/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/GenAiInstrumentation.cs b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/GenAiInstrumentation.cs
index 5b60a9f..e17accd 100644
--- a/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/GenAiInstrumentation.cs
+++ b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/GenAi/GenAiInstrumentation.cs
@@ -90,7 +90,8 @@ public static TResponse Execute<TResponse>(
 
     private static void SetRequestTags(Activity activity, string provider, string operation, string? model)
     {
-        activity.SetTag(SemanticConventions.GenAi.System, provider);
+        // OTel 1.37+: gen_ai.system → gen_ai.provider.name
+        activity.SetTag(SemanticConventions.GenAi.ProviderName, provider);
         activity.SetTag(SemanticConventions.GenAi.OperationName, operation);
 
         if (model is { Length: > 0 })
@@ -108,11 +109,15 @@ private static void SetResponseTags<TResponse>(
         try
         {
             var usage = extractUsage(response);
-            activity.SetTag(SemanticConventions.GenAi.InputTokens, usage.InputTokens);
-            activity.SetTag(SemanticConventions.GenAi.OutputTokens, usage.OutputTokens);
+            // OTel 1.37+: Explicit Usage prefix
+            activity.SetTag(SemanticConventions.GenAi.UsageInputTokens, usage.InputTokens);
+            activity.SetTag(SemanticConventions.GenAi.UsageOutputTokens, usage.OutputTokens);
         }
-        catch
+        catch (Exception ex)
         {
+            // Record extraction failure as event for debugging
+            activity.AddEvent(new ActivityEvent("gen_ai.usage.extraction_failed",
+                tags: new ActivityTagsCollection { ["exception.message"] = ex.Message }));
         }
     }
 
diff --git a/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/SemanticConventions.cs b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/SemanticConventions.cs
index 71c43eb..70bbfe4 100644
--- a/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/SemanticConventions.cs
+++ b/eng/ANcpSdk.AspNetCore.ServiceDefaults/Instrumentation/SemanticConventions.cs
@@ -9,33 +9,131 @@ namespace ANcpSdk.AspNetCore.ServiceDefaults.Instrumentation;
 internal static class SemanticConventions
 {
     /// <summary>
-    ///     GenAI semantic conventions.
+    ///     GenAI semantic conventions (OTel 1.39).
     /// </summary>
     /// <remarks>
     ///     See: https://opentelemetry.io/docs/specs/semconv/gen-ai/
     /// </remarks>
     public static class GenAi
     {
-        /// <summary>The GenAI system (e.g., "openai", "anthropic", "ollama").</summary>
-        public const string System = "gen_ai.system";
+        /// <summary>Schema URL for OTel 1.39 semantic conventions.</summary>
+        public const string SchemaUrl = "https://opentelemetry.io/schemas/1.39.0";
 
-        /// <summary>The operation name (e.g., "chat", "embeddings").</summary>
+        // =====================================================================
+        // CORE ATTRIBUTES (Required/Conditionally Required)
+        // =====================================================================
+
+        /// <summary>gen_ai.provider.name - The GenAI provider (e.g., "openai", "anthropic").</summary>
+        /// <remarks>OTel 1.37+: Replaces deprecated gen_ai.system</remarks>
+        public const string ProviderName = "gen_ai.provider.name";
+
+        /// <summary>gen_ai.operation.name - The operation being performed (e.g., "chat", "embeddings").</summary>
         public const string OperationName = "gen_ai.operation.name";
 
-        /// <summary>The model requested (e.g., "gpt-4o", "claude-3-opus").</summary>
+        // =====================================================================
+        // REQUEST PARAMETERS
+        // =====================================================================
+
+        /// <summary>gen_ai.request.model - The model requested (e.g., "gpt-4o", "claude-3-opus").</summary>
         public const string RequestModel = "gen_ai.request.model";
 
-        /// <summary>The model that generated the response.</summary>
+        /// <summary>gen_ai.request.temperature - Temperature setting (0.0-2.0).</summary>
+        public const string RequestTemperature = "gen_ai.request.temperature";
+
+        /// <summary>gen_ai.request.max_tokens - Maximum tokens to generate.</summary>
+        public const string RequestMaxTokens = "gen_ai.request.max_tokens";
+
+        /// <summary>gen_ai.request.top_p - Nucleus sampling threshold.</summary>
+        public const string RequestTopP = "gen_ai.request.top_p";
+
+        /// <summary>gen_ai.request.top_k - Top-k sampling parameter.</summary>
+        public const string RequestTopK = "gen_ai.request.top_k";
+
+        /// <summary>gen_ai.request.stop_sequences - Stop sequence array.</summary>
+        public const string RequestStopSequences = "gen_ai.request.stop_sequences";
+
+        /// <summary>gen_ai.request.frequency_penalty - Frequency penalty (-2.0 to 2.0).</summary>
+        public const string RequestFrequencyPenalty = "gen_ai.request.frequency_penalty";
+
+        /// <summary>gen_ai.request.presence_penalty - Presence penalty (-2.0 to 2.0).</summary>
+        public const string RequestPresencePenalty = "gen_ai.request.presence_penalty";
+
+        /// <summary>gen_ai.request.seed - Reproducibility seed.</summary>
+        public const string RequestSeed = "gen_ai.request.seed";
+
+        // =====================================================================
+        // RESPONSE ATTRIBUTES
+        // =====================================================================
+
+        /// <summary>gen_ai.response.model - The model that generated the response.</summary>
         public const string ResponseModel = "gen_ai.response.model";
 
-        /// <summary>Number of tokens in the input/prompt.</summary>
-        public const string InputTokens = "gen_ai.usage.input_tokens";
+        /// <summary>gen_ai.response.id - Unique completion identifier.</summary>
+        public const string ResponseId = "gen_ai.response.id";
+
+        /// <summary>gen_ai.response.finish_reasons - Reasons the model stopped generating.</summary>
+        public const string ResponseFinishReasons = "gen_ai.response.finish_reasons";
+
+        // =====================================================================
+        // USAGE/TOKENS (OTel 1.37+ naming)
+        // =====================================================================
+
+        /// <summary>gen_ai.usage.input_tokens - Number of tokens in the input/prompt.</summary>
+        public const string UsageInputTokens = "gen_ai.usage.input_tokens";
+
+        /// <summary>gen_ai.usage.output_tokens - Number of tokens in the output/completion.</summary>
+        public const string UsageOutputTokens = "gen_ai.usage.output_tokens";
+
+        /// <summary>gen_ai.usage.input_tokens.cached - Cached input tokens (Anthropic).</summary>
+        public const string UsageInputTokensCached = "gen_ai.usage.input_tokens.cached";
+
+        /// <summary>gen_ai.usage.output_tokens.reasoning - Reasoning output tokens (o1-style models).</summary>
+        public const string UsageOutputTokensReasoning = "gen_ai.usage.output_tokens.reasoning";
+
+        // =====================================================================
+        // AGENT ATTRIBUTES (OTel 1.38+)
+        // =====================================================================
+
+        /// <summary>gen_ai.agent.id - Unique agent identifier.</summary>
+        public const string AgentId = "gen_ai.agent.id";
+
+        /// <summary>gen_ai.agent.name - Human-readable agent name.</summary>
+        public const string AgentName = "gen_ai.agent.name";
+
+        /// <summary>gen_ai.agent.description - Agent description.</summary>
+        public const string AgentDescription = "gen_ai.agent.description";
+
+        // =====================================================================
+        // TOOL ATTRIBUTES (OTel 1.39)
+        // =====================================================================
+
+        /// <summary>gen_ai.tool.name - Tool name.</summary>
+        public const string ToolName = "gen_ai.tool.name";
+
+        /// <summary>gen_ai.tool.call.id - Tool call identifier.</summary>
+        public const string ToolCallId = "gen_ai.tool.call.id";
+
+        /// <summary>gen_ai.conversation.id - Session/thread identifier.</summary>
+        public const string ConversationId = "gen_ai.conversation.id";
+
+        // =====================================================================
+        // DEPRECATED (for backward compatibility)
+        // =====================================================================
+
+        /// <summary>
+        ///     Deprecated attribute names (pre-1.37). Use for migration/normalization.
+        /// </summary>
+        public static class Deprecated
+        {
+            /// <summary>gen_ai.system - DEPRECATED: Use ProviderName instead.</summary>
+            public const string System = "gen_ai.system";
 
-        /// <summary>Number of tokens in the output/completion.</summary>
-        public const string OutputTokens = "gen_ai.usage.output_tokens";
+            /// <summary>gen_ai.usage.prompt_tokens - DEPRECATED: Use UsageInputTokens instead.</summary>
+            public const string UsagePromptTokens = "gen_ai.usage.prompt_tokens";
 
-        /// <summary>Reasons the model stopped generating (e.g., "stop", "length").</summary>
-        public const string FinishReasons = "gen_ai.response.finish_reasons";
+            /// <summary>gen_ai.usage.completion_tokens - DEPRECATED: Use UsageOutputTokens instead.</summary>
+            public const string UsageCompletionTokens = "gen_ai.usage.completion_tokens";
+        }
     }
 
     /// <summary>