Moderate comments automatically
W2Shield reads every comment on your Instagram posts as it arrives, decides what it is, and hides the ones that make your posts worse — abuse, harassment, scams, and optionally spam. It can also reply to the genuine questions.
The point is that you stop seeing the comments you would only have deleted.
Before you start
- A connected Instagram account — see Connect your Instagram account.
- Funds in your balance. Classification is charged per comment, and with an empty balance W2Shield stops rather than working for free.
Steps
- Open AI Assistants and create a W2Shield.
- Attach it to an account — the Instagram account it should watch.
- Set what it hides, what it replies to, and its limits (all below).
- Enable it. It starts on the next comment that arrives.
What it hides
Abuse and harassment always. Beyond that there is one choice worth understanding:
Also hide spam — promotional and bot-like comments, however polite. This catches the "Great post! Check out my page" comments that are not offensive but are not from real customers either. Most businesses want this on; leave it off if your community genuinely posts links to each other.
Overall strictness decides how borderline a comment has to be before it acts. Start at the middle setting and adjust after reading a week of the activity log — moving straight to the strictest setting tends to hide people who were being sarcastic rather than abusive.
What it replies to
W2Shield can answer genuine questions as well as hide the bad comments.
- Send as — a private direct message or a public reply. Private is the safer default: a wrong public answer is visible to everyone, a wrong private one is visible to one person.
- Wording — your own wording, which sends the same message every time, or AI writes each reply, which answers the actual question using your knowledge base.
- Agent name — the name it signs off with, so people know who they are talking to.
Give it a knowledge base if you want it answering real questions. Its answers come from those documents and nothing else — the same rule as a Support Assistant, and the same advice applies: plain question-and-answer documents work far better than a brochure.
The limits, and why they matter
Four caps, and each one is protecting you from a different failure:
| Setting | What it protects against |
|---|---|
| Max comments moderated per day | A post going viral and costing you a fortune overnight |
| Replies per day | The assistant answering more than you would want it to |
| Per person / 24h | One person getting a stream of automated replies |
| Minimum confidence | It acting on comments it is not sure about |
Set the daily cap deliberately
It is the safety net for the day a post takes off. Past the cap, comments are simply left alone — nothing is hidden and nothing is charged. That is the right failure: it stops spending, it does not start deleting.
What it costs
One charge per comment it classifies, at the AI rate on the pricing page. It appears in your ledger as its own line so you can see it apart from your other AI spend.
Hiding a comment is free. So is skipping one — comments that never reach the AI because it is disabled, the daily cap is reached, the comment is your own, or the balance is empty, cost nothing at all.
A reply it sends is charged as an Instagram message on top, like any other DM. See Reply to Instagram DMs.
Watch what it does
Moderation activity lists every comment it looked at, what it decided and why.
Read it daily for the first week. It is the only way to tell whether the strictness is right, and every wrong decision points at either a setting or a gap in the knowledge base. After that, a weekly skim is enough.
Turning it off
Disable it, or detach it from the account. Either stops it immediately. Comments it already hid stay hidden — unhide those from Comments if you want them back.
If something goes wrong
- It is not doing anything. Check it is enabled and attached to the right account, the account is connected, and the balance is not empty.
- It hid a comment it should not have. Unhide it from Comments and lower the strictness a step.
- It is missing obvious abuse. Raise the strictness, and check the comment's language is in your primary or additional languages.
- It stopped partway through the day. The daily cap. Raise it if the volume is genuine rather than a one-off.
- Its replies are wrong. The knowledge base is wrong, thin, or out of date. Fix the documents, not the wording. Or switch Wording to your own text, which never improvises.