diff --git a/CHANGELOG.md b/CHANGELOG.md
index 75b8f79..a367b75 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,7 +5,13 @@ Follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) · versioned wi
 
 ---
 
-## [Unreleased]
+## [1.26.11] — 2026-09-10
+
+### Added
+- **Comprehensive OCR Element & Grounding Recognition.** Full support for layout-aware vision OCR models (Baidu Unlimited-OCR, DeepSeek-OCR, DeepSeek-OCR-2, PaddleOCR).
+  - Explicit element parsing for Headlines, Paragraphs, Sidenotes/Sidebars, Page Numbers/Folios, Images/Figures, Graphics/Charts/Maps/Treemaps/Diagrams, Tables, and Code blocks.
+  - Native support for grounding coordinate tokens (`<|ref|>...<|/ref|>`, `<|det|>...<|/det|>`, single/double brackets, normalized `0..1000`/`0..1` scale, and pixel space).
+- **Ingestion Preview & Visual Thumbnail Cards.** The live ingestion preview formats non-text visual regions (`[CHART]`, `[MAP]`, `[TABLE]`, `[SIDENOTE]`, `[CODE]`) with clean badges instead of blank gaps, and the magazine review sidebar renders interactive cropped image thumbnails (`mag-block-thumb`) for detected visual figures.
 
 ---
 
diff --git a/VERSION b/VERSION
index 63d92c7..fabf663 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-1.26.10
+1.26.11
diff --git a/core/magazine.py b/core/magazine.py
index 6b378e4..3b78933 100644
--- a/core/magazine.py
+++ b/core/magazine.py
@@ -738,75 +738,124 @@ def _require_fitz():
 # engine invents) is treated as ambiguous body copy and still goes through
 # classify_blocks() for finer distinctions (byline, deck, ...).
 _VISION_LABEL_TYPES = {
-    "footer": BLOCK_FURNITURE,
-    "header": BLOCK_FURNITURE,
+    # Headlines / Titles / Headers
+    "title": BLOCK_TITLE,
+    "headline": BLOCK_TITLE,
+    "article_headline": BLOCK_TITLE,
+    "main_title": BLOCK_TITLE,
+    "heading": BLOCK_HEADING,
+    "header": BLOCK_HEADING,
+    "section_heading": BLOCK_HEADING,
+    "subheading": BLOCK_HEADING,
+    "subtitle": BLOCK_HEADING,
+    "deck": BLOCK_HEADING,
+
+    # Paragraphs / Text / Body
+    "paragraph": BLOCK_BODY,
+    "body": BLOCK_BODY,
+    "content": BLOCK_BODY,
+    "list": BLOCK_BODY,
+    "list_item": BLOCK_BODY,
+    "author": BLOCK_BYLINE,
+    "byline": BLOCK_BYLINE,
+
+    # Sidenotes / Sidebars / Quotes / Footnotes
+    "sidenote": BLOCK_SIDEBAR,
+    "sidebar": BLOCK_SIDEBAR,
+    "aside_text": BLOCK_SIDEBAR,
+    "aside": BLOCK_SIDEBAR,
+    "footnote": BLOCK_SIDEBAR,
+    "footnotes": BLOCK_SIDEBAR,
+    "note": BLOCK_SIDEBAR,
+    "commentary": BLOCK_SIDEBAR,
+    "quote": BLOCK_PULLQUOTE,
+    "pullquote": BLOCK_PULLQUOTE,
+
+    # Page numbers / Furniture / Folios / Sources
     "page_number": BLOCK_FURNITURE,
+    "pagenumber": BLOCK_FURNITURE,
+    "page": BLOCK_FURNITURE,
+    "header_text": BLOCK_FURNITURE,
+    "footer": BLOCK_FURNITURE,
+    "running_header": BLOCK_FURNITURE,
+    "running_footer": BLOCK_FURNITURE,
+    "folio": BLOCK_FURNITURE,
+    "head": BLOCK_FURNITURE,
     "source": BLOCK_FURNITURE,
     "sources": BLOCK_FURNITURE,
     "citation": BLOCK_FURNITURE,
     "citations": BLOCK_FURNITURE,
-    "footnote": BLOCK_FURNITURE,
-    "footnotes": BLOCK_FURNITURE,
-    "title": BLOCK_TITLE,
-    "heading": BLOCK_HEADING,
-    "caption": BLOCK_CAPTION,
-    "image_caption": BLOCK_CAPTION,
-    "table_caption": BLOCK_CAPTION,
-    "headline": BLOCK_TITLE,
-    "article_headline": BLOCK_TITLE,
-    "paragraph": BLOCK_BODY,
-    "section_heading": BLOCK_HEADING,
-    "author": BLOCK_BYLINE,
-    "byline": BLOCK_BYLINE,
-    "sidenote": BLOCK_SIDEBAR,
-    "quote": BLOCK_PULLQUOTE,
-    "pullquote": BLOCK_PULLQUOTE,
+
+    # Images / Figures / Photos
     "image": BLOCK_FIGURE,
     "figure": BLOCK_FIGURE,
     "picture": BLOCK_FIGURE,
+    "photo": BLOCK_FIGURE,
+    "illustration": BLOCK_FIGURE,
+    "non_text": BLOCK_FIGURE,
+    "nontext": BLOCK_FIGURE,
+    "drawing": BLOCK_FIGURE,
+
+    # Graphics / Charts / Diagrams / Maps / Treemaps
     "chart": BLOCK_CHART,
     "graphic": BLOCK_CHART,
     "infographic": BLOCK_CHART,
     "statistics": BLOCK_CHART,
     "diagram": BLOCK_CHART,
+    "map": BLOCK_CHART,
+    "treemap": BLOCK_CHART,
+    "plot": BLOCK_CHART,
+    "flowchart": BLOCK_CHART,
+
+    # Tables
     "table": BLOCK_TABLE,
-    "formula": BLOCK_CODE,
-    "list": BLOCK_BODY,
+    "grid": BLOCK_TABLE,
+    "matrix": BLOCK_TABLE,
+
+    # Code / Formulas / Math
     "code": BLOCK_CODE,
     "code_block": BLOCK_CODE,
+    "formula": BLOCK_CODE,
+    "equation": BLOCK_CODE,
+    "math": BLOCK_CODE,
+
+    # Captions
+    "caption": BLOCK_CAPTION,
+    "image_caption": BLOCK_CAPTION,
+    "figure_caption": BLOCK_CAPTION,
+    "table_caption": BLOCK_CAPTION,
+
+    # Ads
     "advertisement": BLOCK_AD,
     "advert": BLOCK_AD,
     "ad": BLOCK_AD,
-    "aside_text": BLOCK_SIDEBAR,
 }
 
 # One label + its four-number bbox, anchoring a region; content runs until
-# the next such marker. Confirmed live against Baidu Unlimited-OCR's actual
-# output shape: "text [241, 90, 588, 249]Battery technology is entering...".
-_VISION_REGION = re.compile(r"(?m)^(\w+)\s*\[\s*(\d+)\s*,\s*(\d+)\s*,\s*(\d+)\s*,\s*(\d+)\s*\]")
-# Native Baidu Unlimited-OCR document-parsing output. The model card emits
-# one marker per detected region, such as
-# ``<|det|>table [0,0,999,999]<|/det|>...``. Its categories are precisely
-# what we need to retain charts/tables/images as visual blocks instead of
-# treating their axis labels as article prose.
+# the next such marker. Tolerates DeepSeek-OCR / Unlimited-OCR / PaddleOCR shapes
+# like "text [241, 90, 588, 249]", "<|ref|>title<|/ref|><|det|>[[x0, y0, x1, y1]]<|/det|>", etc.
+_VISION_REGION = re.compile(
+    r"(?m)"
+    r"(?:<\|ref\|>\s*(?P<ref>[^<]+?)\s*<\|/ref\|>\s*)?"
+    r"(?:<\|det\|>\s*)?"
+    r"(?P<label>[a-zA-Z0-9_\- ]+?)\s*"
+    r"\[\[?\s*(?P<x0>-?\d+(?:\.\d+)?)\s*,\s*(?P<y0>-?\d+(?:\.\d+)?)\s*,\s*(?P<x1>-?\d+(?:\.\d+)?)\s*,\s*(?P<y1>-?\d+(?:\.\d+)?)\s*\]\]?"
+    r"(?:\s*<\|/det\|>)?"
+    r"\s*:?"
+)
+
 _UNLIMITED_OCR_REGION = re.compile(
-    r"(?m)^<\|det\|>\s*(.*?)\s*(?:\[\s*(-?\d+(?:\.\d+)?)\s*,\s*"
-    r"(-?\d+(?:\.\d+)?)\s*,\s*(-?\d+(?:\.\d+)?)\s*,\s*"
-    r"(-?\d+(?:\.\d+)?)\s*\])?\s*<\|/det\|>")
+    r"(?m)"
+    r"(?:<\|ref\|>\s*(?P<ref>[^<]+?)\s*<\|/ref\|>\s*)?"
+    r"<\|det\|>\s*"
+    r"(?P<label>[a-zA-Z0-9_\- ]+?)\s*"
+    r"(?:\[\[?\s*(?P<x0>-?\d+(?:\.\d+)?)\s*,\s*(?P<y0>-?\d+(?:\.\d+)?)\s*,\s*(?P<x1>-?\d+(?:\.\d+)?)\s*,\s*(?P<y1>-?\d+(?:\.\d+)?)\s*\]\]?)?"
+    r"\s*<\|/det\|>"
+)
 
 
 def ocr_quality_score(text: str) -> float:
-    """Rough plausibility that OCR/extracted text is real prose, 0..1.
-
-    Coarse by design: it catches text that is mostly symbol noise (confirmed
-    live on tesseract line-height artifacts -- ". |", "@)X" scored 0.0/0.33)
-    well enough to gate an expensive vision-OCR escalation, but it cannot and
-    is not meant to catch a plausible-looking character substitution inside
-    an otherwise normal sentence (confirmed live: "K | soll Arbeit abnehmen"
-    -- a single misread "KI" -- still scored 0.81, since 7 of its 8 words are
-    genuine German). Catching that class of error needs a dictionary or
-    language-model check, which is out of scope for a cheap heuristic.
-    """
+    """Rough plausibility that OCR/extracted text is real prose, 0..1."""
     text = (text or "").strip()
     if not text:
         return 0.0
@@ -827,17 +876,7 @@ def ocr_quality_score(text: str) -> float:
 def estimate_font_size_from_area(width: float, height: float, text_length: int,
                                  char_width_ratio: float = 0.5,
                                  line_height_ratio: float = 1.2) -> float:
-    """Estimate a representative font size for a paragraph-level box with no
-    per-line data (a vision OCR engine gives one box per paragraph, not per
-    line, unlike tesseract -- see ocr_page()).
-
-    Derived from the standard area relationship for a wrapped paragraph:
-    box_height ~= (text_length / (box_width / (char_width_ratio*size))) *
-    (line_height_ratio*size), which solves to size ~= sqrt(area / (text_length
-    * char_width_ratio * line_height_ratio)). Using total bbox height directly
-    (as if it were one line) would scale with paragraph length instead of
-    type size -- the same bug already fixed for tesseract's line grouping.
-    """
+    """Estimate a representative font size for a paragraph-level box."""
     if width <= 0 or height <= 0 or text_length <= 0:
         return 0.0
     denom = text_length * char_width_ratio * line_height_ratio
@@ -845,20 +884,17 @@ def estimate_font_size_from_area(width: float, height: float, text_length: int,
 
 
 def parse_vision_ocr_regions(raw: str) -> list[dict[str, Any]]:
-    """Parse a vision OCR engine's labeled-region transcription.
-
-    Tolerates a preamble the model sometimes prepends before the first region
-    (confirmed live: instruction-echoing junk, and the literal decoy phrase
-    "[No text detected]" ahead of real output) by only ever anchoring on the
-    "label [x0, y0, x1, y1]" marker -- anything before the first match, or
-    between calls with zero matches, is discarded rather than misread as
-    content.
-    """
+    """Parse a vision OCR engine's labeled-region transcription."""
     matches = list(_VISION_REGION.finditer(raw or ""))
     regions = []
     for i, m in enumerate(matches):
-        label = m.group(1).lower()
-        x0, y0, x1, y1 = (int(m.group(g)) for g in (2, 3, 4, 5))
+        gd = m.groupdict()
+        raw_label = (gd.get("ref") or gd.get("label") or "text").strip().strip("[]")
+        label = raw_label.lower().replace("-", "_").replace(" ", "_")
+        try:
+            x0, y0, x1, y1 = (float(gd[k]) for k in ("x0", "y0", "x1", "y1"))
+        except (ValueError, TypeError):
+            continue
         start = m.end()
         end = matches[i + 1].start() if i + 1 < len(matches) else len(raw or "")
         text = (raw or "")[start:end].strip()
@@ -866,37 +902,33 @@ def parse_vision_ocr_regions(raw: str) -> list[dict[str, Any]]:
             continue
         regions.append({
             "label": label,
-            "bbox": [float(x0), float(y0), float(x1 - x0), float(y1 - y0)],
+            "bbox": [x0, y0, x1 - x0, y1 - y0],
             "text": text,
         })
     return regions
 
 
 def parse_unlimited_ocr_regions(raw: str) -> list[dict[str, Any]]:
-    """Parse Unlimited-OCR's native ``<|det|>`` document-layout markers.
-
-    Bounding boxes are in its 0–999-ish image coordinate space; callers mark
-    them for conversion to the rendered page's pixels before building blocks.
-    A `[Non-Text]` region is a real visual object even when it has no text.
-    """
+    """Parse Unlimited-OCR's native ``<|det|>`` document-layout markers."""
     matches = list(_UNLIMITED_OCR_REGION.finditer(raw or ""))
     regions = []
     for i, match in enumerate(matches):
-        raw_label = (match.group(1) or "").strip().strip("[]")
+        gd = match.groupdict()
+        raw_label = (gd.get("ref") or gd.get("label") or "").strip().strip("[]")
         label = raw_label.lower().replace("-", "_").replace(" ", "_")
-        coords = match.groups()[1:]
-        if len(coords) != 4 or any(v is None for v in coords):
+        try:
+            x0, y0, x1, y1 = (float(gd[k]) for k in ("x0", "y0", "x1", "y1"))
+        except (ValueError, TypeError):
             continue
-        x0, y0, x1, y1 = (float(v) for v in coords)
         if x1 <= x0 or y1 <= y0:
             continue
         start = match.end()
         end = matches[i + 1].start() if i + 1 < len(matches) else len(raw or "")
         text = (raw or "")[start:end].strip()
-        # The model's generic Non-Text label is an image/figure. It must be
-        # retained even empty, so crop_visual_regions can preserve it.
-        if label in {"non_text", "nontext"}:
+        if label in {"non_text", "nontext", "photo", "drawing", "picture"}:
             label = "image"
+        elif label in {"map", "graphic", "infographic", "diagram", "treemap", "statistics"}:
+            label = "chart"
         if not text and _VISION_LABEL_TYPES.get(label) not in VISUAL_TYPES:
             continue
         regions.append({
@@ -980,17 +1012,19 @@ def vision_ocr_page(png_path: Path, base_url: str, model: str,
         base_url += "/v1"
 
     unlimited_ocr = "unlimited-ocr" in model.lower()
-    prompt = "document parsing." if unlimited_ocr else (
+    prompt = "<|grounding|>Convert the document to markdown." if unlimited_ocr else (
         "Recognise the page layout and transcribe visible text verbatim. "
-        "Return one region per headline, paragraph, image, chart, table, "
-        "code box, caption, header, footer or page number. Use the format "
+        "Return one region per headline, title, paragraph, sidenote, sidebar, "
+        "image, figure, chart, graphic, map, diagram, table, code box, "
+        "caption, header, footer, or page number. Use the format "
         "label [x0, y0, x1, y1]transcription with pixel coordinates in the "
         "supplied image, one region starting on each new line. Labels: "
-        "title, paragraph, image, chart, table, code, caption, header, "
-        "footer, page_number, aside_text. Enclose a whole chart or table "
-        "in one box, not separate axis labels or cells. Preserve table "
-        "rows and columns as Markdown and code indentation and line breaks. "
-        "Images may have empty transcription; do not invent descriptions. "
+        "title, heading, paragraph, sidenote, sidebar, image, figure, chart, "
+        "graphic, map, table, code, caption, header, footer, page_number. "
+        "Enclose a whole chart, diagram, map, graphic, or table in one box, "
+        "not separate axis labels or cells. Preserve table rows and columns "
+        "as Markdown and code indentation and line breaks. Images, figures, and "
+        "charts may have empty transcription; do not invent descriptions. "
         "Do not merge neighbouring articles or include commentary."
     )
     payload = {
diff --git a/routes/magazine.py b/routes/magazine.py
index b0d31b6..e7f9f56 100644
--- a/routes/magazine.py
+++ b/routes/magazine.py
@@ -134,7 +134,17 @@ def _ingest_sync(doc_id: str, job_id: str, ocr_lang: str, render_dpi: int,
             # both wrong for this. Best-effort -- a write failure must never
             # abort ingestion.
             try:
-                preview_text = " ".join(b.text for b in blocks).strip()[:4000]
+                lines = []
+                for b in blocks:
+                    b_type = getattr(b, "type", "")
+                    label = getattr(b, "region_label", None) or b_type or "region"
+                    txt = (getattr(b, "text", "") or "").strip()
+                    if b_type in mag.VISUAL_TYPES or b_type in (mag.BLOCK_FURNITURE, mag.BLOCK_SIDEBAR, mag.BLOCK_PULLQUOTE, mag.BLOCK_AD):
+                        line_str = f"[{label.upper()}] {txt}".strip()
+                        lines.append(line_str)
+                    elif txt:
+                        lines.append(txt)
+                preview_text = "\n\n".join(lines)[:4000]
                 (d / "ingest_preview.json").write_text(
                     json.dumps({"page": page, "text": preview_text}, ensure_ascii=False),
                     "utf-8")
diff --git a/static/index.html b/static/index.html
index 489d706..c9acda3 100644
--- a/static/index.html
+++ b/static/index.html
@@ -10,7 +10,7 @@
   <meta name="format-detection" content="telephone=no">
   <meta name="color-scheme" content="light dark">
   <meta name="theme-color" content="#2563EB">
-<meta name="app-version" content="1.26.10">
+<meta name="app-version" content="1.26.11">
   <link rel="manifest" href="/manifest.webmanifest">
   <link rel="icon" href="/static/icon.svg" type="image/svg+xml">
   <link rel="apple-touch-icon" href="/static/icon.svg">
@@ -27,12 +27,12 @@
 
   <!-- ── Core styles (local — no CDN dependency for first paint) ────────── -->
   <link rel="stylesheet" href="/static/vendor/mdi/materialdesignicons.min.css">
-<link rel="stylesheet" href="/static/style.css?v=1.26.10">
-<link rel="stylesheet" href="/static/css/resource-diagnostics.css?v=1.26.10">
-<link rel="stylesheet" href="/static/css/background-jobs.css?v=1.26.10">
-<link rel="stylesheet" href="/static/css/voice-design.css?v=1.26.10">
-<link rel="stylesheet" href="/static/css/rehearser.css?v=1.26.10">
-<link rel="stylesheet" href="/static/css/audiobook.css?v=1.26.10">
+<link rel="stylesheet" href="/static/style.css?v=1.26.11">
+<link rel="stylesheet" href="/static/css/resource-diagnostics.css?v=1.26.11">
+<link rel="stylesheet" href="/static/css/background-jobs.css?v=1.26.11">
+<link rel="stylesheet" href="/static/css/voice-design.css?v=1.26.11">
+<link rel="stylesheet" href="/static/css/rehearser.css?v=1.26.11">
+<link rel="stylesheet" href="/static/css/audiobook.css?v=1.26.11">
 
 
   <!-- ── Flag icons — non-blocking (loaded async, icons appear after JS) ── -->
diff --git a/static/js/magazine.js b/static/js/magazine.js
index b499eab..e7bf8e7 100644
--- a/static/js/magazine.js
+++ b/static/js/magazine.js
@@ -682,29 +682,35 @@ function renderReview() {
   const nextPage = spanIdx >= 0 && spanIdx < state.reviewSpan.length - 1 ? state.reviewSpan[spanIdx + 1] : null;
   const issuePages = Number(state.layout?.page_count || 0);
   const previousIssuePage = page > 1 ? page - 1 : null;
-  const nextIssuePage = page < issuePages ? page + 1 : null;
+  if (!host) return;
+  const page = state.reviewPage;
+  const pageMeta = page != null ? (state.layout.pages || []).find((p) => p.page === page) : null;
+  const article = state.article;
+  const issuePages = (state.layout || {}).page_count || '—';
+  const nextIssuePage = page != null && page < (state.layout || {}).page_count ? page + 1 : null;
+  const prevIssuePage = page != null && page > 1 ? page - 1 : null;
 
-  let html =
-    '<div class="mag-review-head">' +
-      '<label>Title<span class="mag-review-field-control"><input id="mag-rev-title" type="text" value="' + esc(a.title) + '"><button type="button" class="btn-sm" onclick="magReviewToggleDraw(\'title\')"><span class="mdi mdi-vector-square-edit"></span> Pick from page</button></span></label>' +
-      '<label>Author<span class="mag-review-field-control"><input id="mag-rev-author" type="text" value="' + esc(a.author) + '"><button type="button" class="btn-sm" onclick="magReviewToggleDraw(\'author\')"><span class="mdi mdi-vector-square-edit"></span> Pick from page</button></span></label>' +
-      '<div class="mag-pill ' + conf.cls + '">' + esc(conf.text) + '</div>' +
-    '</div>' +
-    '<p class="muted">Untick anything that is not part of this article — a sidebar, an advert, ' +
-    'or a neighbouring piece. Only ticked text is read aloud. Click a numbered region on the page, ' +
-    'or its matching entry on the right, to line the two up. A region labelled with another ' +
-    'article belongs to a different piece sharing this page. Draw a missing text or image region below.</p>';
+  const included = new Set(article ? article.blocks || [] : []);
+  const ownImages = new Set(article ? article.images || [] : []);
 
+  const pageBlocks = (state.layout.blocks || []).filter((b) => b.page === page);
+  const pageImages = (state.layout.images || []).filter((i) => i.page === page);
+  const regions = [
+    ...pageBlocks.map((b) => ({ kind: 'block', id: b.id, data: b, ro: b.reading_order ?? 0 })),
+    ...pageImages.map((i) => ({ kind: 'image', id: i.id, data: i, ro: i.reading_order ?? 0 })),
+  ].sort((a, b) => a.ro - b.ro);
+
+  let html = '';
   html += '<div class="mag-review-pagenav">' +
-    '<div class="mag-review-page-controls" role="group" aria-label="Review pages">' +
-    '<button type="button" class="btn-sm" aria-label="Previous page" ' + (prevPage ? '' : 'disabled') +
-      ' onclick="magReviewGoPage(' + prevPage + ')"><span class="mdi mdi-chevron-left" aria-hidden="true"></span></button>' +
+      '<div class="mag-review-pagenav-group" role="group" aria-label="Review pages">' +
+      '<button type="button" class="btn-sm" aria-label="Previous page" ' + (state.reviewPageIdx > 0 ? '' : 'disabled') +
+      ' onclick="magReviewStepPage(-1)"><span class="mdi mdi-chevron-left" aria-hidden="true"></span></button>' +
       '<strong class="mag-review-page-number">Page ' + (page ?? '—') + '</strong>' +
-      '<button type="button" class="btn-sm" aria-label="Next page" ' + (nextPage ? '' : 'disabled') +
-      ' onclick="magReviewGoPage(' + nextPage + ')"><span class="mdi mdi-chevron-right" aria-hidden="true"></span></button></div>' +
-    '<div class="mag-review-page-controls mag-review-issue-controls" role="group" aria-label="Whole magazine pages">' +
-      '<button type="button" class="btn-sm" aria-label="Previous magazine page" ' + (previousIssuePage ? '' : 'disabled') +
-      ' onclick="magReviewGoMagazinePage(' + previousIssuePage + ')"><span class="mdi mdi-chevron-left" aria-hidden="true"></span></button>' +
+      '<button type="button" class="btn-sm" aria-label="Next page" ' + (state.reviewPageIdx < (article?.pages || []).length - 1 ? '' : 'disabled') +
+      ' onclick="magReviewStepPage(1)"><span class="mdi mdi-chevron-right" aria-hidden="true"></span></button></div>' +
+      '<div class="mag-review-pagenav-group" role="group" aria-label="Whole magazine pages">' +
+      '<button type="button" class="btn-sm" aria-label="Previous magazine page" ' + (prevIssuePage ? '' : 'disabled') +
+      ' onclick="magReviewGoMagazinePage(' + prevIssuePage + ')"><span class="mdi mdi-chevron-left" aria-hidden="true"></span></button>' +
       '<strong class="mag-review-page-number">Magazine ' + (page ?? '—') + ' / ' + issuePages + '</strong>' +
       '<button type="button" class="btn-sm" aria-label="Next magazine page" ' + (nextIssuePage ? '' : 'disabled') +
       ' onclick="magReviewGoMagazinePage(' + nextIssuePage + ')"><span class="mdi mdi-chevron-right" aria-hidden="true"></span></button></div>' +
@@ -726,7 +732,20 @@ function renderReview() {
     '<button type="button" class="btn-sm" id="mag-reocr-btn" onclick="magReocrPage()">' +
       '<span class="mdi mdi-text-recognition"></span> OCR again (this page)</button>' +
     '<span id="mag-reocr-status" class="muted"></span>' +
-  '</div><div id="mag-draw-hint" role="status"></div>';
+    '</div>';
+
+  html += '<div class="mag-review-fields">';
+  const a = state.article || {};
+  html += '<div class="mag-review-field-row">' +
+      '<label>Title<span class="mag-review-field-control"><input id="mag-rev-title" type="text" value="' + esc(a.title) + '"><button type="button" class="btn-sm" onclick="magReviewToggleDraw(\'title\')"><span class="mdi mdi-vector-square-edit"></span> Pick from page</button></span></label>' +
+      '<label>Author<span class="mag-review-field-control"><input id="mag-rev-author" type="text" value="' + esc(a.author) + '"><button type="button" class="btn-sm" onclick="magReviewToggleDraw(\'author\')"><span class="mdi mdi-vector-square-edit"></span> Pick from page</button></span></label>' +
+    '</div>' +
+    '<span id="mag-draw-hint" class="muted"></span>';
+  html += '</div>';
+
+  const conf = confidenceLabel(article ? article.confidence : 1.0);
+  html += '<span class="mag-review-conf ' + conf.cls + '">' + esc(conf.text) + '</span>';
+  html += '<p class="muted">Untick anything that is not part of this article &mdash; a sidebar, an advert, or a neighbouring piece. Only ticked text is read aloud. Click a numbered region on the page, or its matching entry on the right, to line the two up. A region labelled with another article belongs to a different piece sharing this page. Draw a missing text or image region below.</p>';
 
   if (state.layout.semantic_review === 'partial_failure') {
     html += '<p class="muted" role="status">Some regions could not be grouped automatically. Check paragraph boundaries before approving.</p>';
@@ -739,7 +758,6 @@ function renderReview() {
 
   html += '<div class="mag-review-split">';
 
-  // Left: the rendered page with a numbered box per region.
   html += '<div class="mag-review-page">';
   if (page != null) {
     html += '<div class="mag-review-page-wrap" id="mag-review-page-wrap">' +
@@ -770,7 +788,6 @@ function renderReview() {
   }
   html += '</div>';
 
-  // Right: the same regions as text/figures, numbered to match.
   html += '<div class="mag-blocks">';
   regions.forEach((r, i) => {
     const isImage = r.kind === 'image';
@@ -778,13 +795,18 @@ function renderReview() {
     const icon = magBlockIcon(isImage ? r.data.type || 'image' : r.data.region_label === 'page_number' ? 'page_number' : r.data.type);
     if (isImage) {
       const on = ownImages.has(r.id);
+      const imgFilename = r.data.filename || (r.id + '.png');
+      const imgUrl = '/api/magazine/' + state.doc.id + '/images/' + esc(imgFilename);
       html += '<div class="mag-block mag-block-image' + (on ? '' : ' is-aux') + '" data-block-row="' + esc(r.id) + '" ' +
         'onmouseenter="magReviewHighlightBox(\'' + esc(r.id) + '\')" onmouseleave="magReviewHighlightBox(null)">' +
         '<span class="mag-block-num">' + (i + 1) + '</span>' +
         '<input type="checkbox" aria-label="Include image in playback" ' + (on ? 'checked' : '') + ' onchange="magReviewToggleImage(\'' + esc(r.id) + '\', this.checked)">' +
         '<span class="mag-block-type"><span class="mdi ' + icon + '"></span> ' + esc(r.data.type || 'image') + '</span>' +
-        '<span class="mag-block-text">' + (owner ? 'Belongs to Article ' + owner.index + ': ' + esc(owner.title)
-          : on ? 'Shown during playback for this article.' : 'Not linked to any article on this page.') + '</span>' +
+        '<div class="mag-block-thumb-wrap">' +
+          '<img class="mag-block-thumb" src="' + imgUrl + '" alt="' + esc(r.data.type || 'figure') + '" onerror="this.style.display=\'none\'">' +
+          '<span class="mag-block-text">' + (owner ? 'Belongs to Article ' + owner.index + ': ' + esc(owner.title)
+            : on ? 'Shown during playback for this article.' : 'Not linked to any article on this page.') + '</span>' +
+        '</div>' +
       '</div>';
       return;
     }
diff --git a/tests/test_magazine.py b/tests/test_magazine.py
index a149d5a..6489f88 100644
--- a/tests/test_magazine.py
+++ b/tests/test_magazine.py
@@ -833,7 +833,7 @@ class VisionOcrPageCallTests(unittest.TestCase):
         self.assertEqual(payload["ngram_size"], 35)
         self.assertEqual(payload["window_size"], 128)
         self.assertEqual(payload["images_config"], {"image_mode": "gundam"})
-        self.assertEqual(payload["messages"][0]["content"][0]["text"], "document parsing.")
+        self.assertEqual(payload["messages"][0]["content"][0]["text"], "<|grounding|>Convert the document to markdown.")
 
     def test_other_vision_engines_do_not_receive_unlimited_ocr_options(self):
         body = {"choices": [{"message": {"content":