problem 01
Ghost CMS spam
Fake members, spam comments, and junk submitted through a custom application or contact form on a Ghost publication. The goal is usually a link, a fake account, or a free trial.
use case · ghost cms
Updated November 19, 2026
Ghost spam usually means fake member signups or comment spam on a Ghost publication. Ghost(Pro) blocks signup patterns across its network, and every Ghost site can block email domains under Settings → Membership → Spam filters; native comments are member-only and moderatable. Add server-side content scoring only when a custom form, application, or third-party comment tool collects free text that Ghost's built-in controls do not evaluate.
The word "ghost" covers two different problems: spam inside Ghost CMS, and fake referral traffic that shows up in analytics. This guide handles both, then shows where a content classifier earns its place.
problem 01
Fake members, spam comments, and junk submitted through a custom application or contact form on a Ghost publication. The goal is usually a link, a fake account, or a free trial.
problem 02
Visits in Google Analytics from domains that never sent you a real reader. It is not a Ghost bug and not a form problem. Filter it by hostname; it does not need a spam API. Read the analytics explainer.
native first
Before adding a vendor, use the controls you already pay for. Ghost(Pro) monitors signups across its managed hosting network and blocks patterns of abuse automatically. On any Ghost site, an admin can add abusive email domains to the membership spam filter. Native comments require a logged-in member and can be held for moderation, which removes the anonymous drive-by comment entirely.
Source: Ghost's signup spam protection documentation and comments documentation, checked 7 November 2026.
cleanup
Filter the member list by signup date and source. Accounts that never opened an email, share a disposable domain, or arrived in a burst are the first to review. Export the list before deleting so you have a record.
Add each abusive domain to Settings → Membership → Spam filters. One domain per line. This stops the same addresses from signing up again without affecting legitimate members on normal providers.
Ghost comments are member-only, so moderation is a quality control rather than a bot wall. Remove link-only replies and generic AI-written praise, then keep moderation on for new members while trusted members post freely.
If the complaint is fake traffic in Google Analytics, filter by hostname instead of installing a form-protection product. The two problems share a word, not a fix.
defence layers
| Layer | Catches | Misses | Reader friction |
|---|---|---|---|
| Ghost(Pro) network protection | Signup patterns Ghost sees across its hosting network. | Self-hosted sites, and spam that looks like a normal signup. | None |
| Email domain blocklist | Known throwaway or abusive domains you add by hand. | New domains, compromised inboxes, and content spam. | None |
| Member-only comments + moderation | Drive-by anonymous comments; lets you hold the rest. | Spam posted by a registered member, and custom forms. | Low |
| Honeypot field | Simple bots that fill every input. | Headless browsers and human spam farms. | None if implemented accessibly |
| Content scoring API | Low-quality text in applications, bios, messages, and custom forms. | Email reputation, IP reputation, and browser behaviour. | None |
| CAPTCHA or challenge | Automated sessions at the edge. | Human spam and motivated headless browsers. | High |
The order matters: block known domains first, keep comments member-only, then add content scoring only where free text enters your workflow. A CAPTCHA should be the exception for a flood, not the default for every reader.
shortlist
Ghost gives you more than most publishers realise. This is the order to reach for them, and the one case where a classifier is worth adding.
Add abusive email domains under Settings → Membership → Spam filters. Ghost(Pro) also blocks signup patterns across its network.
Require a logged-in member to comment and hold new members' comments for review. Removes anonymous drive-by spam entirely.
Score the free text in a custom application, contact form, or third-party comment tool before it creates a member or sends mail.
Put an invisible challenge in front of the signup or form endpoint when you are actually being flooded.
Ghost's native behaviour is documented in Ghost's help centre and checked 19 November 2026. Pricing for the API layer is Siftfy's published pricing.
where an api fits
Ghost's native controls evaluate an email domain and a membership state. They do not read the free-text answer in a custom application, the bio on a member profile, or the message in a third-party contact form. That is the gap a content classifier fills: one server-side call returns a calibrated 0–1 spam probability, and your application decides whether to allow, review, or drop the submission.
// Cloudflare Worker in front of a custom Ghost application form.
// Ghost's Admin API creates the member only after the text check passes.
export default {
async fetch(request, env) {
const { name, email, application } = await request.json();
const scored = await fetch("https://api.siftfy.io/v1/predict", {
method: "POST",
headers: {
"Content-Type": "application/json",
"X-API-Key": env.SIFTFY_KEY,
},
body: JSON.stringify({ text: application }),
signal: AbortSignal.timeout(2000),
}).then((r) => (r.ok ? r.json() : null)).catch(() => null);
// Fail open on transport failure. Review the uncertain band; block only
// above the model's own confidence ceiling.
const spam = scored && scored.max_confidence !== null
&& scored.spam_probability > scored.max_confidence;
if (spam) return Response.json({ ok: true }); // silent drop, no signal
const res = await fetch(`${env.GHOST_API_URL}/ghost/api/admin/members/`, {
method: "POST",
headers: {
"Content-Type": "application/json",
Authorization: `Ghost ${env.GHOST_ADMIN_KEY}`,
},
body: JSON.stringify({ members: [{ name, email, labels: ["application"] }] }),
});
return Response.json({ ok: res.ok }, { status: res.ok ? 200 : 502 });
},
};Keep the email, IP, and browser checks in the systems that already own them. Siftfy evaluates language, not identity. If you only need a pre-built snippet, the Ghost spam filter example walks through the same flow in more detail. If the free text arrives through a third-party comment tool rather than a custom form, the honeypot guide covers the cheap first layer, and the Akismet alternatives comparison covers the comment-filter market.
decision rule
No custom free text on the site
Use Ghost's domain filter and comment moderation. Do not buy a classifier.
A custom application or contact form collects a message
Score the message server-side before it creates a member or sends an email.
Third-party comments or a headless front end
Score the body before publication; keep the platform's moderation queue as the review band.
A bot flood, not a content problem
Add rate limits and a challenge at the edge. A text classifier is the wrong layer for volume.
Start with Ghost's own controls: the membership spam filter blocks abusive email domains, Ghost(Pro) blocks signup patterns across its network, and comments are member-only and moderatable. Add an API-based content classifier only if a custom application, contact form, or third-party comment tool collects free text that Ghost cannot read. A CAPTCHA is a fallback for a volume flood, not a default filter.
Ghost(Pro) monitors signups across its hosting network and blocks patterns it identifies as abuse, and every Ghost site can block specific email domains under Settings → Membership → Spam filters. Native comments are member-only and can be moderated. Self-hosted Ghost does not get the network-level signup protection, so pair the domain blocklist with rate limits or a content check if you collect free text.
Filter the member list by signup date or source, identify accounts that never opened an email and share an obvious disposable domain, then delete or unsubscribe them in bulk. Add the abusive domains to Settings → Membership → Spam filters so the same addresses cannot sign up again. Export the list first if you want an audit trail.
The phrase also describes fake referral traffic in Google Analytics: visits from spam domains such as buttons-for-website or semalt that never reached your site. That is a different problem from Ghost CMS spam. Filter those referrals by hostname; they do not represent real readers and do not need a form-protection product.
Yes. Ghost comments are limited to logged-in members, so the first defence is member signup quality. Keep moderation on, block the email domains you see repeatedly, and add server-side content scoring if you run a custom application form or a third-party comment tool. A CAPTCHA is a fallback for volumetric abuse, not the default for every reader.
It is designed for untrusted user text, not for your own writing. Do not score your own posts or newsletter copy; promotional writing is exactly what the model is trained to flag. Use it on application answers, profile bios, contact messages, and comments submitted by members.
start free
10,000 requests a month free, no card. One endpoint, a calibrated probability, and no puzzle for your readers.