diff --git a/.github/workflows/devlens-live-integration.yml b/.github/workflows/devlens-live-integration.yml
index ea52b28..070409b 100644
--- a/.github/workflows/devlens-live-integration.yml
+++ b/.github/workflows/devlens-live-integration.yml
@@ -66,7 +66,7 @@ jobs:
           assert set(report["scores"]) == dimensions
           assert all(0 <= int(v) <= 100 for v in report["scores"].values())
           assert os.environ["BADGE"].startswith("https://img.shields.io/badge/")
-          assert report["action_version"] == "2.0.8"
+          assert report["action_version"] == "2.0.9"
 
           print(f"Live DevLens integration passed: {score}/100")
 
diff --git a/CHANGELOG.md b/CHANGELOG.md
index c3fe757..3a199b6 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -4,6 +4,10 @@ All notable changes to DevLens, newest first.
 
 ## Unreleased — 2026-09-30
 
+- Corrected Issue Response scoring to exclude pull requests from issue counts.
+- Added deterministic regression fixtures for issue-response scoring.
+- Updated live integration expectations for the `2.0.9` candidate.
+
 ### Documentation and distribution status
 - Verified by the Human Owner that the live GitHub Marketplace listing displays **DevLens Repo Health** as **Latest v2.0.8** with the current Action-only product content.
 - Closed the Marketplace publication/verification blocker and advanced the project to Phase 2 — Adoption and onboarding.
diff --git a/docs/AI_PROJECT_CONTEXT.md b/docs/AI_PROJECT_CONTEXT.md
index dfa7192..3cc1158 100644
--- a/docs/AI_PROJECT_CONTEXT.md
+++ b/docs/AI_PROJECT_CONTEXT.md
@@ -310,7 +310,7 @@ As of 2026-09-30:
 
 - `v2.0.0` exists as the original Marketplace release tag.
 - `v2.0.1` was previously published successfully.
-- `v2.0.8` is the latest verified release and was created and published by the automatic critical-change release workflow.
+- `v2.0.8` is the latest currently published release; `v2.0.9` is the pending candidate for the issue-response scoring correction.
 - The release preflight executed the real Action successfully and validated the reported Action version against the release version.
 - The floating `v2` tag was updated successfully to `v2.0.8`.
 - The automatic release workflow detects critical production paths on pushes to `main`, derives the release version from `scripts/devlens.py`, runs static validation and live Action preflight, updates `v2`, and publishes the GitHub Release.
@@ -331,10 +331,10 @@ As of 2026-09-30:
 - Vercel configuration is removed from the repository.
 - Legacy scanner code is removed.
 - Legacy Groq scoring dependencies/inputs are removed.
-- README, contributor, security, documentation, and changelog materials are aligned with v2.0.8.
+- README, contributor, security, documentation, and changelog materials are aligned with the current v2.0.9 candidate where applicable.
 - Live GitHub integration has executed the actual Action successfully.
 - README persistence has been tested on a disposable branch and verified through GitHub API.
-- Current release line is v2; `v2.0.8` is the latest verified version.
+- Current release line is v2; `v2.0.8` is the latest verified published version, with `v2.0.9` pending validation.
 - Floating `v2` points to the latest verified v2 release (`v2.0.8`).
 - Repository release automation has been hardened so manual releases complete in one workflow run and do not depend on a token-generated tag push triggering a second workflow.
 - Repository metadata may still contain stale external/homepage information and should be verified before treating metadata cleanup as complete.
@@ -389,7 +389,7 @@ Remaining Phase 1 operational check:
 
 ### Phase 3 — Scoring quality and observability — NEXT
 
-- Expand automated edge-case tests.
+- **IN PROGRESS:** Expand automated edge-case tests, beginning with Issue Response.
 - Improve evidence shown for individual dimension scores.
 - Reduce unnecessary API calls and startup cost.
 - Improve sparse-history behavior.
@@ -498,7 +498,7 @@ The repository adopts the AI CEO/CIO → COO execution model and treats document
 
 ### D-012 — Action report version must match implementation
 **Status:** DECISION / ACTIVE.  
-The machine-readable report action version must identify the actual DevLens implementation version. This is enforced by release preflight and live integration validation. The current verified implementation reports `2.0.8`.
+The machine-readable report action version must identify the actual DevLens implementation version. This is enforced by release preflight and live integration validation. The current verified implementation candidate reports `2.0.9`; the production release remains `v2.0.8` until release preflight completes.
 
 ### D-010 — Automated semantic v2 release flow
 **Status:** DECISION / ACTIVE.  
diff --git a/scripts/devlens.py b/scripts/devlens.py
index c7713a9..9dcffc3 100644
--- a/scripts/devlens.py
+++ b/scripts/devlens.py
@@ -81,15 +81,25 @@ def score_ci():
         return 0
     except: return 0
 
+def score_issue_response(open_count, closed_count):
+    # Keep issue response deterministic and independent of pull-request closure activity.
+    """Score actual GitHub issues only; pull requests are excluded from both counts."""
+    if open_count < 0 or closed_count < 0:
+        raise ValueError("Issue counts cannot be negative")
+    total = open_count + closed_count
+    if total == 0:
+        return 100
+    return int(closed_count / total * 100)
+
 def score_issues():
     try:
-        open_i   = repo.open_issues_count
-        closed_i = list(repo.get_issues(state="closed"))[:50]
-        if not closed_i and open_i == 0: return 100
-        total = open_i + len(closed_i)
-        if total == 0: return 100
-        return int(len(closed_i) / total * 100)
-    except: return 50
+        # GitHub's /issues collection includes pull requests. Search with is:issue
+        # so the dimension measures issue response rather than PR closure activity.
+        open_i = g.search_issues(query=f"repo:{REPO_NAME} is:issue is:open").totalCount
+        closed_i = g.search_issues(query=f"repo:{REPO_NAME} is:issue is:closed").totalCount
+        return score_issue_response(open_i, closed_i)
+    except Exception:
+        return 50
 
 def score_community():
     return min(int(math.log1p(repo.stargazers_count)*15)+int(math.log1p(repo.forks_count)*10), 100)
@@ -162,7 +172,7 @@ def dim_bar(score):
 badge_url = (f"https://img.shields.io/badge/DevLens%20Health-{health}%2F100"
              f"-{badge_color(health)}?style={BADGE_STYLE}&logo=github")
 
-report = {"repo":REPO_NAME,"score_model":"action-v2-9d","action_version":"2.0.8","health_score":health,"scores":scores,
+report = {"repo":REPO_NAME,"score_model":"action-v2-9d","action_version":"2.0.9","health_score":health,"scores":scores,
           "badge_url":badge_url,"generated_at":now.isoformat()}
 
 print(json.dumps(report, indent=2))
diff --git a/tests/test_issue_response_scoring.py b/tests/test_issue_response_scoring.py
new file mode 100644
index 0000000..858f14f
--- /dev/null
+++ b/tests/test_issue_response_scoring.py
@@ -0,0 +1,31 @@
+import ast
+from pathlib import Path
+
+source = Path("scripts/devlens.py").read_text()
+tree = ast.parse(source)
+fn = next(node for node in tree.body if isinstance(node, ast.FunctionDef) and node.name == "score_issue_response")
+module = ast.Module(body=[fn], type_ignores=[])
+namespace = {}
+exec(compile(module, "scripts/devlens.py", "exec"), namespace)
+
+score = namespace["score_issue_response"]
+cases = [
+    ((0, 0), 100),
+    ((0, 10), 100),
+    ((10, 0), 0),
+    ((5, 5), 50),
+    ((1, 9), 90),
+]
+for args, expected in cases:
+    actual = score(*args)
+    assert actual == expected, (args, actual, expected)
+
+for args in [(-1, 0), (0, -1)]:
+    try:
+        score(*args)
+    except ValueError:
+        pass
+    else:
+        raise AssertionError(f"negative counts accepted: {args}")
+
+print("Issue-response scoring fixtures passed.")