Question 1

How does the AI decide between Hide, Reply, Keep, and Escalate?

Accepted Answer

The model classifies each comment by intent: harmless social engagement (KEEP), real customer questions or feedback worth a public reply (REPLY), spam / scam / hateful low-effort noise (HIDE), and serious complaints, threats, or PR-risk content that needs a human (ESCALATE). It considers tone, content, sentiment, and red-flag patterns.

Question 2

How accurate is it?

Accepted Answer

On most batches it agrees with experienced human moderators around 85–90% of the time. Edge cases — sarcasm, in-jokes, niche slang, ambiguous complaints — are where you should always sanity-check before taking destructive actions like blocking users.

Question 3

Can I paste comments from any platform?

Accepted Answer

Yes. The classifier is platform-agnostic. Paste comments from Instagram, TikTok, Facebook, YouTube, Threads, Twitter/X, or any other source — one per line, up to 50 per run.

Question 4

Does FeedGuardians do this automatically on my real comments?

Accepted Answer

Yes. FeedGuardians connects to your Instagram, Facebook, TikTok, and YouTube accounts and runs this exact triage on every new comment in real time — automatically hiding spam, queuing replies, and escalating sensitive ones to your team.

Question 5

Is my data stored?

Accepted Answer

No. Comments are sent to OpenAI for classification and not stored on our servers. Don't paste personally identifiable data — replace customer names with placeholders if needed.

Question 6

How is this different from Meta's built-in filter?

Accepted Answer

Meta's filter is keyword matching only — it can't tell a joke from a threat or a real question from spam. This tool understands context, sarcasm, and intent.

Commentaire AI Triage

Ceci mais sur chaque commentaire, en temps réel

FAQ

Détecteur de Spam & Arnaques

Générateur de Mots Cachés

Analyseur de Sentiment en Masse