diff --git a/scripts/document_gen/templates/mcs150_pdf_filler.py b/scripts/document_gen/templates/mcs150_pdf_filler.py
index f52936e..d3a72ed 100644
--- a/scripts/document_gen/templates/mcs150_pdf_filler.py
+++ b/scripts/document_gen/templates/mcs150_pdf_filler.py
@@ -209,23 +209,19 @@ def fill_mcs150(intake: dict, order_number: str = "") -> str:
     checkbox_on["certifyBox"] = True
 
     # ── Apply fields to PDF ──────────────────────────────────────────
-    # Update text fields
-    writer.update_page_form_field_values(
-        writer.pages[0],
-        {k: v for k, v in field_updates.items() if v},
-        auto_regenerate=False,
-    )
-
-    # For multi-page forms, try updating all pages
-    for page_idx in range(len(writer.pages)):
+    # Apply text-field values to every page. The fields live on pages 9-11 of
+    # the official form, so we update across all pages (pypdf silently ignores
+    # field names not present on a given page). auto_regenerate=True makes pypdf
+    # build appearance streams from the values, so viewers that ignore
+    # /NeedAppearances (Preview, Chrome) still render the text.
+    text_values = {k: v for k, v in field_updates.items() if v}
+    for page in writer.pages:
         try:
             writer.update_page_form_field_values(
-                writer.pages[page_idx],
-                {k: v for k, v in field_updates.items() if v},
-                auto_regenerate=False,
+                page, text_values, auto_regenerate=True,
             )
-        except Exception:
-            pass
+        except Exception as exc:
+            LOG.debug("Form field apply on page failed: %s", exc)
 
     # Apply checkbox fields
     for field_name, checked in checkbox_on.items():
diff --git a/scripts/document_gen/templates/verify_mcs150.py b/scripts/document_gen/templates/verify_mcs150.py
new file mode 100644
index 0000000..34fc3af
--- /dev/null
+++ b/scripts/document_gen/templates/verify_mcs150.py
@@ -0,0 +1,112 @@
+"""Verify an MCS-150 PDF actually RENDERS the expected values.
+
+OCR tooling (tesseract/pdftoppm/fitz) is not available in the workers
+container, so instead of rasterising we verify what a viewer would draw by
+reading each text widget's *appearance stream* (/AP /N), which is the literal
+content the PDF renders. After fill_mcs150 with auto_regenerate=True, pypdf
+writes these streams, so their presence + content is ground truth that the
+value will be visible (not just sitting in /V under a blank widget).
+
+Usage:
+    python3 -m scripts.document_gen.templates.verify_mcs150 [DOT_NUMBER]
+
+Exits non-zero if any expected value is missing from the rendered output.
+"""
+from __future__ import annotations
+
+import re
+import sys
+
+from pypdf import PdfReader
+
+from .mcs150_pdf_filler import fill_mcs150
+
+
+def _appearance_text(reader: PdfReader) -> dict[str, str]:
+    """Return {field_name: rendered text from its /AP /N stream}."""
+    out: dict[str, str] = {}
+    for page in reader.pages:
+        annots = page.get("/Annots")
+        if not annots:
+            continue
+        for ref in annots:
+            obj = ref.get_object()
+            name = obj.get("/T")
+            if name is None:
+                continue
+            ap = obj.get("/AP")
+            text = ""
+            if ap:
+                n = ap.get_object().get("/N")
+                if n is not None:
+                    try:
+                        raw = n.get_object().get_data()
+                        # Tj / TJ text operators carry the drawn glyphs.
+                        for m in re.findall(rb"\((.*?)\)\s*Tj", raw):
+                            text += m.decode("latin-1", "ignore")
+                        for m in re.findall(rb"\[(.*?)\]\s*TJ", raw):
+                            for s in re.findall(rb"\((.*?)\)", m):
+                                text += s.decode("latin-1", "ignore")
+                    except Exception:
+                        pass
+            out[str(name)] = text.strip()
+    return out
+
+
+def main() -> int:
+    dot = sys.argv[1] if len(sys.argv) > 1 else "1609564"
+
+    # Simulate enrichment by pulling the FMCSA census the same way the handler
+    # does, then fill.
+    from scripts.workers.services.mcs150_update import MCS150UpdateService
+
+    svc = MCS150UpdateService()
+    census = svc._fetch_carrier_record(dot)
+    if not census:
+        print(f"FAIL: no FMCSA census for DOT {dot}")
+        return 2
+
+    intake = dict(census)
+    intake.setdefault("signer_name", "Test Signer")
+    intake.setdefault("signer_title", "Owner")
+
+    path = fill_mcs150(intake, order_number="CO-VERIFY")
+    reader = PdfReader(path)
+    rendered = _appearance_text(reader)
+
+    # What we expect to be visible on the rendered form.
+    checks = {
+        "1bizName": intake.get("legal_name", ""),
+        "3principalStreet": intake.get("address_street", ""),
+        "4principalCity": intake.get("address_city", ""),
+        "5principalState": intake.get("address_state", ""),
+        "6principalZip": intake.get("address_zip", ""),
+        "16usdotNumber": intake.get("dot_number", ""),
+        "19irsNumber": intake.get("ein", ""),
+        "certifyName": intake.get("signer_name", ""),
+    }
+
+    print(f"=== MCS-150 render verification for DOT {dot} ===")
+    print(f"PDF: {path}")
+    print(f"census fields: {sorted(census.keys())}\n")
+
+    failures = []
+    for field, expected in checks.items():
+        if not expected:
+            continue
+        got = rendered.get(field, "")
+        ok = expected.upper().replace(" ", "") in got.upper().replace(" ", "")
+        flag = "OK " if ok else "MISS"
+        print(f"  [{flag}] {field}: expected '{expected}' | rendered '{got}'")
+        if not ok:
+            failures.append(field)
+
+    if failures:
+        print(f"\nFAIL: {len(failures)} field(s) not rendered: {failures}")
+        return 1
+    print("\nPASS: all expected values render in their appearance streams.")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())