diff --git a/README.md b/README.md
index de6f038..e44c450 100644
--- a/README.md
+++ b/README.md
@@ -109,6 +109,17 @@ Your data is stored in `~/.tinyweb/`:
 
 This allows your data to persist between upgrades and stay separate from the application.
 
+### Backups
+
+Back up the whole `~/.tinyweb/` directory periodically. The two files that matter:
+
+- **`tinyweb_identity`** is your permanent mesh identity. If you lose it, your destination hash changes and every subscriber has to re-subscribe to the new one. Keep it somewhere you trust; the file is `0600` by default.
+- **`index.db`** is your full reading history — every page, note, tag, and synced remote page. Losing it loses everything you've curated.
+
+`models/` and `index.hnsw` are re-derivable (the model will re-download, and the HNSW index rebuilds from the database on next startup with semantic search enabled) so they don't need to be backed up.
+
+The `/export` page produces a JSON dump of your pages. It's a migration aid — it doesn't preserve your identity file, your custom template, or subscription state. A full restore needs a copy of `~/.tinyweb/`.
+
 ### Docker
 
 Data is stored in the `/data` volume inside the container. Use a volume mount to persist data:
diff --git a/handlers.py b/handlers.py
index 47f38e1..e47520b 100644
--- a/handlers.py
+++ b/handlers.py
@@ -334,6 +334,18 @@ def handle_search(query):
     sub_count = ""
     if q and remote_rows:
         sub_count = f" + {len(remote_rows)} from subscriptions"
+    welcome_html = ""
+    if count == 0 and not q:
+        welcome_html = (
+            '<section style="margin-top:1.5rem;max-width:40em">'
+            '<p>Your index is empty.</p>'
+            '<p>tinyweb is a personal search engine for pages you save. '
+            'The index stays on your machine; so does every search.</p>'
+            '<p>From here: <a href="/add">add a page</a>, '
+            '<a href="/style">get the bookmarklet</a>, or '
+            '<a href="/subscriptions">subscribe to another instance</a>.</p>'
+            '</section>'
+        )
     return _respond(
         f'<form method="get" action="/">'
         f'<input name="q" value="{esc(q)}" placeholder="search your index" size="40">'
@@ -341,6 +353,7 @@ def handle_search(query):
         f'</form>'
         f'<p class="meta">{count} pages indexed'
         f' · <a href="/add">+ add url</a></p>'
+        f'{welcome_html}'
         f'{result_html}'
         f'{_page_nav(page, total_results, f"/?q={esc(q)}") if q else ""}'
         f'{trusted_html}{remote_html}'
@@ -547,6 +560,43 @@ def handle_pages(query=None):
     )
 
 
+def _render_bulk_delete_confirm(page_ids):
+    """Server-side confirmation page for bulk deletion — mirrors handle_delete_confirm."""
+    db = get_db()
+    try:
+        placeholders = ",".join("?" * len(page_ids))
+        rows = db.execute(
+            f"SELECT id, url, title FROM pages WHERE id IN ({placeholders})",
+            page_ids,
+        ).fetchall()
+    finally:
+        return_db(db)
+    if not rows:
+        return _redirect("/pages")
+    items = "".join(
+        f'<li><b>{esc(r["title"] or r["url"])}</b><br>'
+        f'<small>{esc(r["url"])}</small></li>'
+        for r in rows
+    )
+    hidden_ids = "".join(
+        f'<input type="hidden" name="ids" value="{int(r["id"])}">' for r in rows
+    )
+    n = len(rows)
+    return _respond(
+        f"<h1>confirm delete</h1>"
+        f"<p>Remove the following {n} page{'' if n == 1 else 's'}?</p>"
+        f"<ul>{items}</ul>"
+        f'<form method="post" action="/pages/bulk">'
+        f'{_csrf_field()}'
+        f'{hidden_ids}'
+        f'<input type="hidden" name="action" value="delete">'
+        f'<input type="hidden" name="confirmed" value="1">'
+        f'<button type="submit">yes, delete {n} page{"" if n == 1 else "s"}</button>'
+        f"</form>"
+        f' <a href="/pages">cancel</a>'
+    )
+
+
 def handle_bulk_action(body):
     ids = body.get("ids", [])
     action = body.get("action", [""])[0]
@@ -557,6 +607,10 @@ def handle_bulk_action(body):
         page_ids = [int(i) for i in ids]
     except ValueError:
         return _error(400)
+    # Require an explicit second-step confirmation for bulk delete — the JS
+    # confirm() on /pages is a first-line filter only.
+    if action == "delete" and body.get("confirmed", [""])[0] != "1":
+        return _render_bulk_delete_confirm(page_ids)
     db = get_db()
     try:
         if action == "delete":
@@ -871,7 +925,8 @@ def handle_style_form(msg=""):
         f"<p>Drag this link to your bookmarks bar. Click it on any page to index it instantly.</p>"
         f'<p><a href="javascript:void(fetch(\'http://localhost:8080/bookmark?url=\'+encodeURIComponent(location.href)+\'&token={_get_bookmark_token()}\').then(r=>r.text()).then(t=>alert(t)).catch(()=>alert(\'tinyweb not running\')))">+ save to {esc(name)}</a></p>'
         f"<h2>reset</h2>"
-        f'<form method="post" action="/style/reset">'
+        f'<form method="post" action="/style/reset" '
+        f'onsubmit="return confirm(\'Reset the template to default? Your custom template will be lost.\')">'
         f'{_csrf_field()}'
         f'<button type="submit">reset template to default</button>'
         f"</form>"
@@ -960,6 +1015,22 @@ def handle_about():
         f'</ul>'
         f'{sharing_html}'
         f'{hash_html}'
+        f'<h2>your data</h2>'
+        f'<p>Everything is stored locally under <code>~/.tinyweb/</code>:</p>'
+        f'<ul>'
+        f'<li><code>tinyweb_identity</code> — your permanent mesh identity. '
+        f'If you lose this file, your destination hash changes and subscribers '
+        f'have to re-subscribe to the new one.</li>'
+        f'<li><code>index.db</code> — your full reading history: every page, '
+        f'note, tag, and synced remote page.</li>'
+        f'<li><code>models/</code> — the semantic search model if you enabled it '
+        f'(redownloadable, safe to delete).</li>'
+        f'</ul>'
+        f'<p><b>Back up <code>~/.tinyweb/</code> periodically.</b> '
+        f'Copying the whole directory to another device preserves your identity and index together. '
+        f'The <a href="/export">export</a> page gives you a JSON dump of pages only — '
+        f'it does not preserve your identity or subscription state, so it is a migration aid, '
+        f'not a substitute for a full backup.</p>'
         f'<h2>what is the slow web?</h2>'
         f'<p>The slow web is a movement for intentionality over speed, '
         f'human curation over algorithmic feeds, privacy over surveillance, '