<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[Vestifina Studio]]></title><description><![CDATA[Vestifina Studio]]></description><link>https://vestifinastudio.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Sat, 21 Nov 2026 18:17:10 GMT</lastBuildDate><atom:link href="https://vestifinastudio.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[Making a static site citable by AI search: JSON-LD, llms.txt, AI-crawler robots rules and IndexNow on Cloudflare Pages]]></title><description><![CDATA[A working checklist with copy-paste snippets, from a real no-framework site.]]></description><link>https://vestifinastudio.hashnode.dev/making-a-static-site-citable-by-ai-search-json-ld-llms-txt-ai-crawler-robots-rules-and-indexnow-on-cloudflare-pages</link><guid isPermaLink="true">https://vestifinastudio.hashnode.dev/making-a-static-site-citable-by-ai-search-json-ld-llms-txt-ai-crawler-robots-rules-and-indexnow-on-cloudflare-pages</guid><category><![CDATA[SEO]]></category><category><![CDATA[Web Development]]></category><category><![CDATA[cloudflare]]></category><category><![CDATA[structured data]]></category><dc:creator><![CDATA[Vestifina Studio]]></dc:creator><pubDate>Mon, 09 Nov 2026 12:33:50 GMT</pubDate><content:encoded><![CDATA[<p>"Optimise for ChatGPT / Perplexity / AI Overviews" gets sold as something new. For a static site it's mostly old-fashioned technical SEO done carefully, plus a few small files. This post walks through what we actually shipped on a plain HTML site hosted on Cloudflare Pages (no framework, no build step), with the snippets, and is honest about which parts matter and which are speculative.</p>
<p>Live example: <a href="https://vestifina.com/">vestifina.com</a> — a small shop selling document templates. View source on any product or guide page to see everything below in place.</p>
<h2>0. What "citable" actually means</h2>
<p>AI answer engines mostly retrieve from a normal search index (Bing for ChatGPT search and Copilot, Google's own index for AI Overviews, their own crawlers for Perplexity and others), then quote passages. So the requirements are:</p>
<ol>
<li>The page is <strong>crawlable and indexable</strong> by the relevant bots.</li>
<li>The page has <strong>extractable facts</strong>: a clear H1, question-style H2s, a short summary near the top, explicit prices/specs.</li>
<li><strong>Structured data matches the visible text</strong> (mismatches are a spam signal, not a boost).</li>
<li>The index <strong>learns about changes quickly</strong>.</li>
</ol>
<p>Google's own guidance says AI features need nothing beyond normal SEO best practice. Keep that in mind when someone sells you a magic file.</p>
<h2>1. Put the answer in the HTML, near the top</h2>
<p>Before any markup, fix the content. Each product and guide page got a short "at a glance" block right under the H1:</p>
<pre><code class="language-html">&lt;section class="glance" aria-labelledby="glance-h"&gt;
  &lt;h2 id="glance-h"&gt;What is the Day-of timeline?&lt;/h2&gt;
  &lt;dl&gt;
    &lt;div&gt;&lt;dt&gt;What it is&lt;/dt&gt;&lt;dd&gt;A one-page photographer run-of-show with travel buffers…&lt;/dd&gt;&lt;/div&gt;
    &lt;div&gt;&lt;dt&gt;Who it is for&lt;/dt&gt;&lt;dd&gt;Wedding photographers who…&lt;/dd&gt;&lt;/div&gt;
    &lt;div&gt;&lt;dt&gt;What you get&lt;/dt&gt;&lt;dd&gt;Editable Word (.docx) and PDF files, A4 and US Letter.&lt;/dd&gt;&lt;/div&gt;
    &lt;div&gt;&lt;dt&gt;Price&lt;/dt&gt;&lt;dd&gt;$24 USD, one-time payment.&lt;/dd&gt;&lt;/div&gt;
  &lt;/dl&gt;
&lt;/section&gt;
</code></pre>
<p>A <code>&lt;dl&gt;</code> is semantic, short, and trivially quotable. It also surfaced a real bug for us: older copy described the files as "Editable HTML", which was wrong. Writing a fact-only summary forces you to find contradictions like that.</p>
<p>Rename vague H2s ("Good to know") to questions people actually ask ("What format are the files?").</p>
<h2>2. JSON-LD that mirrors the page</h2>
<h3>Product + Offer</h3>
<pre><code class="language-html">&lt;script type="application/ld+json"&gt;
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "6-hour and 8-hour wedding day timeline presets",
  "description": "Four ready wedding day timelines: 6 hours and 8 hours of coverage, each with and without a first look…",
  "brand": { "@type": "Brand", "name": "Vestifina" },
  "sku": "VF-TIMELINE-PRESETS",
  "url": "https://vestifina.com/timeline-presets/",
  "image": "https://vestifina.com/img/og-timeline-presets.png",
  "category": "Digital templates for wedding photographers",
  "audience": { "@type": "Audience", "audienceType": "Wedding photographers" },
  "offers": {
    "@type": "Offer",
    "price": "15.00",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "url": "https://vestifina.com/timeline-presets/",
    "seller": { "@type": "Organization", "name": "Vestifina", "url": "https://vestifina.com/" }
  }
}
&lt;/script&gt;
</code></pre>
<p>Deliberately <strong>not</strong> included: <code>aggregateRating</code> and <code>review</code>. The shop is new and has no reviews; inventing them violates Google's structured-data policies and is the fastest way to get manual action. If you have real reviews later, add them then.</p>
<h3>Article + BreadcrumbList + FAQPage on guides</h3>
<pre><code class="language-html">&lt;script type="application/ld+json"&gt;
{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[
  {"@type":"Question","name":"Is 6 hours of wedding photography enough?",
   "acceptedAnswer":{"@type":"Answer","text":"For many single-venue weddings, yes, if…"}}
]}
&lt;/script&gt;
</code></pre>
<p>Note: since 2023 Google shows FAQ rich results only for a small set of authoritative sites, so don't expect a SERP feature. We keep FAQPage because it's valid, cheap, and other consumers read it — but the <strong>visible</strong> <code>&lt;details&gt;&lt;summary&gt;</code> FAQ is what matters.</p>
<h3>Keep markup and text identical: generate both from one source</h3>
<p>The easiest way to stop drift is to render the visible FAQ and the JSON-LD from the same data:</p>
<pre><code class="language-python">import json, html

faq = [("Is 6 hours enough?", "For many single-venue weddings, yes…"),
       ("When should coverage start?", "About two hours before the ceremony…")]

visible = "".join(
    f'&lt;details&gt;&lt;summary&gt;{html.escape(q)}&lt;/summary&gt;'
    f'&lt;div class="ans"&gt;&lt;p&gt;{html.escape(a)}&lt;/p&gt;&lt;/div&gt;&lt;/details&gt;' for q, a in faq)

ld = {"@context": "https://schema.org", "@type": "FAQPage",
      "mainEntity": [{"@type": "Question", "name": q,
                      "acceptedAnswer": {"@type": "Answer", "text": a}} for q, a in faq]}
script = f'&lt;script type="application/ld+json"&gt;{json.dumps(ld, ensure_ascii=False)}&lt;/script&gt;'
</code></pre>
<p>And a CI-style check that fails if they ever diverge:</p>
<pre><code class="language-python">import re, json, html, glob, sys
bad = 0
for f in glob.glob("site/**/index.html", recursive=True):
    s = open(f).read()
    lds = [json.loads(x) for x in re.findall(
        r'&lt;script type="application/ld\+json"&gt;(.*?)&lt;/script&gt;', s, re.S)]  # raises on invalid JSON
    shown = [html.unescape(x) for x in re.findall(r"&lt;summary&gt;(.*?)&lt;/summary&gt;", s)]
    marked = [q["name"] for l in lds if l.get("@type") == "FAQPage" for q in l["mainEntity"]]
    if marked and shown != marked:
        print("FAQ mismatch:", f); bad += 1
sys.exit(bad)
</code></pre>
<h2>3. robots.txt: allow AI <em>search</em> bots explicitly</h2>
<p>There are two families of AI crawlers: <strong>training</strong> crawlers (GPTBot, Google-Extended, ClaudeBot…) and <strong>search/retrieval</strong> agents (OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot…). You can treat them differently. We allow the search ones explicitly:</p>
<pre><code class="language-txt">User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Bingbot
Allow: /

User-agent: *
Allow: /

Sitemap: https://example.com/sitemap.xml
</code></pre>
<p><strong>Cloudflare gotcha:</strong> if the zone has Cloudflare's managed robots.txt / "block AI training bots" setting on, Cloudflare <strong>prepends</strong> its own block (content signals + Disallow rules for training bots) to whatever file you deploy. <code>curl</code> your live <code>/robots.txt</code> — it won't match your repo file. That's fine if you only want to block training, but know it's there, and make sure it doesn't block the search agents you care about.</p>
<h2>4. llms.txt (cheap, speculative)</h2>
<p><a href="https://llmstxt.org/">llms.txt</a> is a proposed Markdown index for LLMs. Google has said it doesn't use it; some tools and agents do read it. It costs ten minutes, so we ship one with a one-paragraph description, the products with prices, and the key guides:</p>
<pre><code class="language-markdown"># Vestifina

&gt; Editable client paperwork templates for independent wedding photographers: Word (.docx) and PDF…
&gt; Prices in USD, one-time payment. Checkout hosted by Polar (merchant of record).

## Templates
- [Day-of timeline ($24)](https://vestifina.com/day-of-timeline/): photographer run-of-show with travel buffers…

## Free guides
- [6 hour wedding photography timeline](https://vestifina.com/guides/6-hour-wedding-photography-timeline/): two sample schedules…
</code></pre>
<p><strong>Cloudflare Pages gotcha:</strong> if your project has no <code>404.html</code>, Pages treats it as a single-page app and serves <code>/index.html</code> for unknown paths with a <strong>200</strong>. Before we added the file, <code>/llms.txt</code> returned our homepage HTML. Always check new well-known files with <code>curl -I</code> and look at <code>content-type</code>, not just the status code.</p>
<h2>5. Redirect, don't duplicate</h2>
<p>We had an old <code>/blog/</code> section that duplicated newer guides. Instead of <code>noindex</code>, they became permanent redirects in Pages' <code>_redirects</code> file (both with and without trailing slash):</p>
<pre><code class="language-txt">/blog/second-shooter-brief-template   /guides/second-shooter-tips/   301
/blog/second-shooter-brief-template/  /guides/second-shooter-tips/   301
</code></pre>
<h2>6. Tell the index: Bing Webmaster Tools + IndexNow</h2>
<p>ChatGPT search and Copilot lean on Bing's index, so Bing Webmaster Tools is worth the five minutes. Verification on a static site is just a file: download <code>BingSiteAuth.xml</code> from the dashboard, put it in the deploy root, deploy, click verify, submit <code>sitemap.xml</code>. Bing's dashboard also has an AI-performance report showing citations in Copilot/Bing answers — ours shows zero so far, which is the honest starting point for a new site.</p>
<p><a href="https://www.indexnow.org/">IndexNow</a> lets you push changed URLs to participating engines (Bing, Yandex, Seznam, Naver…) instead of waiting for a recrawl. Generate a key, host it as <code>/&lt;key&gt;.txt</code> containing the key, then POST after each deploy:</p>
<pre><code class="language-bash">KEY=your32charkey
curl -s -o /dev/null -w "%{http_code}\n" \
  -H 'Content-Type: application/json; charset=utf-8' \
  -d "{\"host\":\"example.com\",\"key\":\"$KEY\",
       \"keyLocation\":\"https://example.com/$KEY.txt\",
       \"urlList\":[\"https://example.com/guides/new-page/\",
                    \"https://example.com/sitemap.xml\"]}" \
  https://api.indexnow.org/indexnow
# 200 or 202 = accepted
</code></pre>
<p>Only submit URLs that actually changed; this isn't a "ping everything daily" API. (Google doesn't participate in IndexNow, and its Indexing API is restricted to job-posting and livestream pages — don't use it for normal pages. Submit the sitemap in Search Console instead.)</p>
<h2>7. Deploy safely on Cloudflare Pages</h2>
<p>Our deploy loop with wrangler, no Git integration:</p>
<pre><code class="language-bash"># 1. back up what's live (keep the previous deployment id too)
tar czf backup-$(date +%Y%m%d-%H%M).tgz site/
npx wrangler pages deployment list --project-name my-site | head

# 2. preview branch → get a unique *.pages.dev URL to check
npx wrangler pages deploy site --project-name my-site --branch preview

# 3. production
npx wrangler pages deploy site --project-name my-site --branch main
</code></pre>
<p>Then verify production, not just the preview: every sitemap URL returns 200, every JSON-LD block parses, prices in Product markup match the page:</p>
<pre><code class="language-bash">curl -s https://example.com/sitemap.xml | grep -o '&lt;loc&gt;[^&lt;]*' | sed 's/&lt;loc&gt;//' |
while read u; do
  printf '%s %s\n' "$(curl -s -o /dev/null -w '%{http_code}' "$u")" "$u"
done | grep -v '^200' || echo "all 200"
</code></pre>
<p>Rollback is one click in the Pages dashboard (or redeploy the backup folder).</p>
<h2>What we don't know yet</h2>
<p>This is a new site; there are no citations or meaningful traffic numbers to report yet, and anyone promising a specific uplift from these steps is guessing. What these steps do guarantee is that nothing technical stands between a good page and being retrieved and quoted. The content still has to be the best answer to the question.</p>
<hr />
<p><em>Disclosure: the example site belongs to the author's small studio, Vestifina.</em></p>
]]></content:encoded></item></channel></rss>