Skip to content
Vol. VI · Publisher Edition · Austin, Texas Issue No. 24

AI-Assisted Comment Moderation

The fortified home for serious online conversation.

Smarter comments. Stronger communities. Built for publishers who refuse to let their comment sections rot.

Comment Castle is an AI-assisted moderation platform combining real-time spam blocking, toxicity scoring, and reader engagement tools into one fortress — already filtering more than 1.2 billion comments for the brands that take their trust & safety seriously.

Book a 20-Minute Demo Read the editorial brief

Trusted by publishers reaching a combined 180M monthly unique readers · SOC 2 Type II · ISO 27001

Two community managers reviewing a moderated comment thread on a monitor in a quiet newsroom.
Fig. 1 — The publishing floor, after the spam has been cleared.

Audited figures, not marketing fluff — every number below comes from a 2024 deployment census across our 3,400+ active publisher sites.

1.2B+ Comments Moderated Since our founding in 2019 across the full publisher fleet.
3,400+ Active Publisher Sites From daily newsrooms to weekly magazines and vertical communities.
99.2% Average Spam-Block Rate Reported across enterprise deployments in 2024.
71% Cost Reduction in 90 Days Audited average across 280 customer migrations to the platform.

96.4% precision on Toxicity Score v3 · sub-180ms average moderation latency · 612 new publisher accounts onboarded in 2024.

The Core Patent

Toxicity Score v3 — a brand-safe yardstick, not a moral panic button.

Most classifiers hand you a number and a shrug. Toxicity Score v3 ships with a precision claim you can take to your general counsel: 96.4% against brand-safe baselines, trained on 41 million labeled comment samples, returning verdicts in under 180 milliseconds.

The score is calibrated to your publication's house style. A sports vertical, a parenting forum, and a financial newsroom will each receive a different curve — what reads as vigorous disagreement in one section is a flag in another. Our linguists tune the baseline with you during onboarding, then re-validate quarterly.

Every score is auditable. Each flagged comment arrives with the contributing signals surfaced as plain English, so your editors can override, escalate, or release in a single click — and your compliance team can reconstruct any decision six months later.

  • 01 41M labeled comment samples — the largest editorial-grade training set in the category.
  • 02 Sub-180ms moderation latency — fastest in the independent CommunityTools Review Q2 2024 benchmark.
  • 03 Per-publication brand-safe baseline, tuned by our in-house linguists.
  • 04 Plain-English rationale for every score, exportable for compliance audits.
Product plate showing the Toxicity Score v3 moderation dashboard with rationale panel.
Fig. 2 — Toxicity Score v3 in production. The rationale panel is the part your legal team will actually read.

Six Pillars, One Fortress

Every moderation job a publisher has, covered by a single platform.

Comment Castle replaces the six separate tools your team is currently duct-taping together. Each pillar is production-grade on day one, and they share a single moderation queue, a single reputation graph, and a single audit log.

i.

Real-Time Spam Blocking

Link farms, bot bursts, and SEO-bait comments are stopped at the edge — 99.2% caught before a human ever sees the queue. Link-trust reputation propagates across our entire publisher network, so a domain banned on one site is blocked everywhere within minutes.

ii.

Toxicity Scoring & Triage

Every comment receives a Toxicity Score v3 reading with a plain-English rationale. Editors triage the flagged 0.8% instead of re-reading the other 99.2%. Brand-safe baselines are calibrated to your section and re-validated each quarter.

iii.

Thread Ranking & Reading Order

Quality rises. Threads are reordered by contribution, civility, and reader reputation — not by chronology. The result: the comment a reader actually finishes is the comment worth finishing. Average time-on-thread climbs within the first month.

iv.

Reader Reputation Graph

A persistent, portable reputation score for every commenter — earned by the quality of their contributions over time, not by logins or paid tiers. Repeat good-faith posters get lighter friction; serial bad actors get escalated before they ever publish.

v.

Auto-Moderation Rules

Build a rulebook in plain English ("release if score < 0.2 and author reputation > 60") and the platform enforces it 24/7. Every rule ships with an audit trail and a kill-switch — because rules that nobody can revoke are the dangerous kind.

vi.

Engagement Analytics

Weekly and quarterly reports that connect moderation decisions to downstream metrics: time-on-page, returning commenters, subscription conversion. The data your editor-in-chief actually asks for, on a single page they will actually read.

Lives Where Your Content Already Lives

One-click integrations for the CMS your editorial team is already using.

No replatforming. No custom build. Comment Castle ships native plugins for the four CMSs that account for the majority of publisher deployments, plus a documented REST and Webhook API used by 600+ engineering teams to wire moderation into custom stacks.

Average install time on WordPress: 11 minutes · documented reference implementations on GitHub · migration concierge included on every enterprise plan.

The Compliance Posture a Regulated Publisher Requires

Trust is not a badge. It is a posture, an audit trail, and a phone number you can actually call.

Comment Castle was founded in 2019 in Austin, Texas by two former Trust & Safety leads at major social platforms. We build for buyers who have to answer to a general counsel, a DPO, and a CISO — in that order. The four numbers below are the ones those buyers ask us for first.

SOC 2
Type II

Audited Annually

Full Type II report available under NDA. Trust Services Criteria covered: Security, Availability, Confidentiality.

ISO
27001

Certified Information Security

ISMS covers the moderation pipeline end-to-end, including the on-device pseudonymization stage described below.

GDPR
Pseudonymization

On-Device, In Production

First moderation vendor to ship on-device GDPR pseudonymization in production. PII never leaves the reader's browser in identifiable form.

99.97%

Trailing-12-Month Uptime

Public uptime page. Status incidents are posted in plain English, with the customer impact and the fix.

Recognized by Nieman Lab's 2024 roundup of moderation tools for newsrooms · winner of the 2024 SaaS Awards, "AI Innovation in Media" · "Best Comment Moderation Platform" at the Online Publishing Awards in 2022, 2023, and 2024.

Book a Demo

Twenty minutes. One of our engineers. Your actual comment section.

A live, working walkthrough of the Comment Castle moderation platform — using a sample of your publication's comment traffic (anonymized on-device) so you leave the call knowing exactly what the platform would catch on Monday morning.

  • Toxicity Score v3 calibrated to your brand-safe baseline, live on your data.
  • A spam-block rate estimate against your last 30 days of comment traffic.
  • Integration walkthrough for your CMS and your engineering team's stack.
  • Pricing tailored to monthly unique readers and moderation volume — no per-seat surprises.

Prefer email? [email protected] · Or call +1 (512) 555-0184, Monday–Friday, 9am–6pm Central.

Replies within one business day · 1101 East 6th Street, Austin, TX