Aller au contenu principal
AIFORYA

AI Comment Moderation for WordPress: Spam, Toxicity, Trolls

Comments scored by AI (spam, toxicity, threat, sentiment), polite trolls spotted, conflicts predicted, no AI call on public rendering. One-time payment.

AIFORYA Comment Moderator: community moderation by AI, beyond anti-spam

AIFORYA Comment Moderator analyses every comment with multi-axis AI (spam, toxicity, insult, threat, sexual content, off-topic, sentiment) and automates approval, holding or spam-flagging according to your thresholds. Beyond classic anti-spam, it adds a layer competitors don't have: insidious troll detection, conflict prediction, echo chamber detection, image analysis and reply suggestion. You keep full control of your costs with the BYOK model — your own API key, no per-comment commission.

Compatible with WordPress 6.0+, PHP 8.0+, multi-language, and WooCommerce.

See the price and licences

See features

Sold individually for a one-time payment, unlimited AI usage via your own API key.

The invisible burden of manual moderation

A dynamic comment section is a powerful engine for engagement, SEO, and community. Yet, without an effective moderation strategy, it quickly becomes a costly liability, a source of stress, and a risk to your reputation.

Web professionals face a frustrating and quantifiable reality:

  • The overwhelming volume of spam. Independent studies estimate that, on sites without advanced protection, up to 85% of submitted comments are actually spam. This constant deluge drowns out legitimate conversations. It makes finding conversational gems nearly impossible.

  • Insidious toxicity. Hateful messages, harassment, misinformation, or unconstructive criticism degrade brand image and audience trust. Managing such content represents a significant mental and emotional burden. Potential legal risks also exist.

  • The polite troublemaker, invisible to anti-spam. Sealioning, gaslighting, concern trolling or strawmanning are never "spam". They pass every classic filter, yet they rot conversations. No anti-spam detects them.

  • Limitations of classic tools. Traditional solutions quickly show their weaknesses against evolving threats:

    • Keyword filtering: It is rigid and lacks context. It can block relevant comments (false positives) while letting cleverly formulated spam pass through.
    • IP blacklists: These are easily circumvented by bot networks that use thousands of different IP addresses.
    • CAPTCHAs: They introduce friction for legitimate users. They degrade the experience and are increasingly foiled by automated resolution services.

Existing alternatives are excellent anti-spam tools: they block the obvious bot, but stay blind to the register of a message, to toxicity, trolling and conversational health. AIFORYA Comment Moderator sits a notch above — on community health — and can coexist with an existing volume-based anti-spam.

Key features

Here is what the extension actually does, verified in the code, with no empty promises. The AI layer relies on your API key; no call is made during public rendering, so your pages' performance stays intact.

Multi-Axis AI Analysis on Submission

  • Description: For every submitted comment — or in the background every 5 minutes to catch any unscored ones — the AI evaluates the content across seven axes (spam, toxicity, insult, threat, sexual content, off-topic) plus sentiment and detected language. According to thresholds you set, the comment is auto-approved, held, or flagged as spam. Certain signals force an escalation to human review (high threat or sexual content).
  • Concrete benefit: You go beyond a simple "spam / not spam". The extension understands the register of a message and applies your policy automatically, while respecting WordPress's native moderation settings.
  • BYOK integration: The analysis goes through your key (Anthropic by default, an economical model), with no per-call surcharge imposed by AIFORYA.

Insidious Troll Detection

  • Description: The AI analyses a comment in light of its author's history (last ten messages) and identifies five patterns of hard-to-spot trolling: sealioning, gaslighting, concern trolling, whataboutism and strawmanning. It returns a score and quotes the passages that serve as evidence.
  • Concrete benefit: You identify the polite troublemaker that slips past every anti-spam. No WordPress competitor offers this analysis.
  • BYOK integration: The contextual analysis relies on a finer model (Anthropic) via your key.

Conflict Prediction

  • Description: On a discussion thread, the AI reads the last five comments and estimates the probability that a message will trigger an argument — a risk score from 0 to 100 — along with a suggested intervention for the moderator.
  • Concrete benefit: You act before the flare-up, rather than assessing the damage afterwards. Early de-escalation beats a thousand deletions.

Echo Chamber Detection & Divergent Question

  • Description: A local algorithm (Jaccard similarity over the last thirty comments of an article, with no AI call) spots threads turning into echo chambers and ranks the most at-risk articles. On demand, the AI then generates an open-ended question to re-energise the debate.
  • Concrete benefit: You keep a lively and diverse conversation. Detection is free because it's local; the AI only steps in for the relaunch, at your initiative.

Vision Analysis of Attached Images

  • Description: When a comment contains an image, the multimodal AI categorises it (screenshot, meme, promotional spam, personal photo, sensitive content, other) and estimates an NSFW score as well as a visual-spam score.
  • Concrete benefit: You also moderate what isn't text — a classic anti-spam blind spot.
  • BYOK integration: Vision is available via Anthropic, OpenAI or Google. DeepSeek is excluded, with an explicit error message if you try to use it.

AI Reply Suggestion & Publishing

  • Description: The AI drafts a contextual reply based on the article, the comment and the parent thread, in the same language. The "Publish" button creates a genuine approved child comment, signed by the administrator.
  • Concrete benefit: You reply quickly while keeping editorial control: nothing is published without your validation click.

Monthly Community Health Report

  • Description: Each month (via scheduled task or on demand), the extension aggregates volume, statuses, sentiment and echo signals, then the AI produces an actionable markdown report. The history keeps the last twelve reports.
  • Concrete benefit: You get a synthetic read on your community's health without re-reading everything — an ideal deliverable for client reporting.

Author Reputation & Token Economy

  • Description: A 0-to-100 score is computed per email address from actual history: (approved − 2 × spam) / total. A "trusted" author (at least five approved comments, zero spam) is auto-approved with no AI call at all.
  • Concrete benefit: Your regulars pass instantly, and you save tokens on already-proven profiles — a strong argument against volume-billed SaaS.

Classic Moderation Foundation: PII, Blocklist, Bulk & Alerts

  • Local PII detection: A local regex scan spots emails, phone numbers, IBANs, bank cards and identity numbers, and displays an alert badge to the moderator. 100% local processing: no AI call, no storage, and no automatic masking. It's a signal to draw your attention, not a transformation of the comment — we prefer to say so clearly.
  • IP / email / domain blocklist: An exact-match short-circuit, applied before any AI call (so at zero cost), with manual management and automatic purge.
  • Bulk moderation & filters: Checkboxes, batch actions (approve / spam / trash / hold) and a bar of combinable filters (sentiment, AI score range, article, dates).
  • Critical email alert: An email is sent to the configured moderator when a serious signal crosses its threshold (threat, sexual content or insult). The vast majority of traffic generates no notification.

Multi-Provider Personal API Key

  • Description: Connect your API key among four providers — Anthropic, OpenAI, Google and DeepSeek — with a per-provider test button and key-prefix validation that prevents input errors. The service automatically steps up (from a fast model to a finer one) for writing tasks.
  • Concrete benefit: You choose your provider, change your key whenever you want, and pay for your calls at cost, directly with the provider. No content data transits through AIFORYA.

Concrete use cases

The WordPress Freelancer

Sophie manages 15 client sites. Manual moderation took her 4 to 5 hours a week, an unbilled burden that eroded her margin. With multi-axis analysis and author reputation, her regulars pass without any AI call and most sorting happens on its own according to her thresholds. Each month, she sends her clients the community health report — a tangible deliverable that seeded a new high-value "Strategic Monitoring" offering.

The Digital Agency

Pivot agency deploys the solution across its entire portfolio of sites under maintenance contracts. Bulk moderation, troll detection and the monthly report have become its reporting tools. The commercial argument has become powerful: "Your community is protected by 24/7 semantic moderation that spots not just spam, but trolls and debates going off the rails." The profitability of their contracts soared by eliminating hidden labour costs.

The High-Traffic Media Outlet

An online media outlet suffered comment threads that regularly turned into arguments. Conflict prediction flags at-risk messages before the flare-up, echo chamber detection spots articles going in circles, and the divergent question relaunches the debate. The moderation team now intervenes in a targeted way, on the 5% of cases that really matter, instead of re-reading everything.

Transparent and predictable pricing

AIFORYA Comment Moderator is sold on its own, for a one-time payment: no subscription, no automatic renewal, nothing to cancel. You choose between three licences — 1 site, 5 sites or 25 sites — and your purchase includes 12 months of updates. The price of each licence is shown in the purchase box on this page. And if the extension does not suit you, you have a 14-day money-back guarantee: a full refund, simply by writing to support@aiforya.fr. After those 12 months, the extension keeps working, including its AI features with your own key: only the updates stop.

Important note on the BYOK (Bring Your Own Key) model: Your purchase covers the software licence and 12 months of updates. Your AI calls (Anthropic, OpenAI, Google, DeepSeek) are billed separately and directly by the provider you choose, via your own API key. For a standard-traffic site, this cost is generally between 2€ and 8€ per month — and lower still, since author reputation and the blocklist cut a large share of AI calls. AIFORYA applies NO markup on your AI consumption, and NO per-comment commission.

Frequently Asked Questions

General AIFORYA Questions

1. How do I create my Anthropic (or OpenAI, Google, DeepSeek) API key? The process takes less than 5 minutes. Create an account on your chosen provider's platform, add a payment method, then generate a key in the "API Keys" section. It is recommended to create a specific key per site. AIFORYA's documentation provides detailed guides for each step.

2. Is there a subscription to cancel? No. The extension is bought once, with no subscription and no automatic renewal: there is nothing to cancel. Your purchase includes 12 months of updates, and you have a 14-day money-back guarantee — a full refund, simply by writing to support@aiforya.fr.

3. How does the BYOK model work in practice? You paste your own API key into the extension settings. AI calls are then made directly from your WordPress server to the provider, with your key. You pay actual usage at cost, without intermediary or markup, and your content data never transits through AIFORYA.

4. What are the actual AI costs? They are very low because moderation handles little text. An AI bill typically ranges between 2€ and 8€ per month for a medium-traffic site — often less, since trusted authors and the blocklist avoid a large share of calls. You can set budgets and spending alerts directly on your provider's dashboard.

5. Is my data stored by AIFORYA? No. AIFORYA never intercepts, reads, or stores the content of your comments. Data transits encrypted (HTTPS) between your WordPress server and the AI provider you have chosen.

6. Is the extension compatible with Elementor / Divi / Breakdance? Yes. AIFORYA Comment Moderator integrates with the native WordPress comment system (wp_comments). It is therefore compatible with all themes and page builders that respect WordPress development standards.

7. Is there a trial period? No. Free extensions are available on the "Free extensions" page, with no credit card required, and this extension comes with a 14-day money-back guarantee: a full refund on simple request to support@aiforya.fr.

8. What happens if Anthropic changes its prices? The BYOK model guarantees you transparency and responsiveness. If AI model prices drop (a frequent trend), you benefit instantly. If a provider becomes too expensive, you switch to another compatible one by simply changing your key.

9. Can I use the extension on multiple sites? Yes: the extension is available with a 1-site, 5-site or 25-site licence.

10. How can I contact AIFORYA? By writing to support@aiforya.fr — including to request a refund within the 14-day guarantee period.

Specific Questions for Comment Moderator

11. In which languages can the AI moderate comments? The AI is multilingual. Thanks to large models (Anthropic, OpenAI, Google), the extension analyses content whatever its language and automatically detects each comment's language (ISO code). French, English, Spanish, German, Italian, Dutch and many others are supported.

12. Is the extension compatible with WooCommerce product reviews? Yes. WooCommerce product reviews use the standard WordPress comment system. AIFORYA Comment Moderator intercepts and analyses them natively, with the same analysis axes and sentiment.

13. How are AI false positives handled? You always remain in control. Held comments join the native WordPress moderation queue, where you review them and approve or restore them in one click. Each decision feeds the author's reputation score, which refines subsequent decisions.

14. What happens if my API key is invalid or my provider account runs out of credits? The extension is designed to be resilient. In case of a communication problem with the API, the analysis returns an explicit error rather than a false neutral score, and the comment stays in the manual moderation queue. No comment is lost or published without verification.

15. Does AI moderation slow down my site for visitors? No, the impact on front-end performance is zero. No LLM call happens during public rendering: the analysis is triggered when the comment is saved (comment_post) and then in the background (WP-Cron). Page loading is never blocked.

16. Are all the AI functions actually operational? Yes — and we only describe what is. Multi-axis analysis, troll detection, conflict prediction, the divergent question, image analysis, reply suggestion and the monthly report are fully wired and go through your API key. Echo chamber detection and author reputation are local, deterministic algorithms, with no black box. We prefer to announce less and deliver for real.

17. Does PII detection actually mask personal data? To be precise: no. PII detection is a local regex scan that displays an alert badge to the moderator when an email, phone number, IBAN or card number appears in a comment. It performs no automatic masking and stores nothing. It's a signal to help you decide, not an anonymisation of the content — and we will never present it otherwise.

AIFORYA Commitments

Client Sovereignty The BYOK model is a guarantee. You control your data, your costs and your dependencies. AIFORYA provides the tool, you keep the power. Your content data never transits through our servers.

Radical Transparency No hidden markups, no opaque credits, no per-comment commission. And we announce what the product does today — not what it might do one day. When a function is a signal (PII detection) rather than a transformation, we say so.

Service Continuity AIFORYA undertakes, should it permanently cease its activity, to publish the source code of its plugins under a free licence (MIT), so that you can keep using them — a commitment written into its Terms of Sale (Article 11).

GDPR Compliance by Architecture Privacy is built into the architecture: no AI call during public rendering, no content data sent to AIFORYA, 100% local PII detection, data purge on uninstall. Compliance is a prerequisite, not a checkbox.

Uncompromising Quality The software is maintained with regular updates for security and compatibility with the latest WordPress versions, along with continuous feature enrichment.

Free yourself from moderation. Focus on growth.

Stop losing valuable hours filtering out noise. Move from anti-spam to genuine community health: trolls spotted, conflicts anticipated, debates relaunched, costs controlled to the cent. All in BYOK, with no per-comment commission.

Install AIFORYA Comment Moderator. It is bought once, with no subscription, and comes with a 14-day money-back guarantee; free extensions are also available, with no credit card required. Installation and basic configuration take less than 5 minutes.

See the price and licences

Consult technical documentation