Skip to main content
Vid ReceiptsHelp center
FeaturesReceipts

AI moderation

How AI moderation works on Vid Receipts, what gets reviewed, what gets blocked, and what to do if your content is flagged.

Vid Receipts uses AI-powered moderation to help keep public feeds accurate, helpful, and respectful. There are two moderation systems: text moderation (receipt content) and image moderation (file uploads).


Receipt text moderation

When you publish a receipt to a public feed, our AI moderation system automatically reviews the content. The system evaluates your receipt for quality, relevance, and compliance with our Content guidelines.

Moderation happens in the background, you don't need to do anything extra. You can see your receipt while it is reviewed. Other readers see it once it is approved.

Changing displayed text, a quote, a link preview, or an attached media reference sends public and custom-feed content back for review. Approval applies to the version that was reviewed. A delayed result cannot approve a newer edit. Content already held for manual review stays held until an operator reviews it.

An operator can restore a receipt that was rejected by mistake. Restoring it publishes it again. Followers receive a new-receipt notification at most once for that receipt, so restoring an already announced receipt does not send another.

What the AI checks

The moderation system looks at several aspects of your receipt:

  • Relevance: Is the receipt related to the video it's attached to?
  • Accuracy: Does the content appear factual and well-supported? Receipts that fact-check claims need to hold up to scrutiny.
  • Tone: Is the language respectful and constructive?
  • Spam detection: Is the content repetitive, promotional, or low-effort?
  • Policy compliance: Does it follow the platform's content guidelines?

What triggers text moderation

Moderation applies to:

  • Public feed receipts: Automatically moderated after publishing
  • Edits to public receipts: Re-evaluated when you update content
  • Custom feed receipts and comments: Reviewed by default. A feed owner can turn off automatic review only while the feed is neither discoverable nor global.

Moderation does not apply to:

  • Private receipts: Only visible to you, so no moderation needed
  • Draft receipts: Unpublished content is not reviewed

Text review includes displayed quote and link-preview text. It does not inspect the contents of a linked website or replace image and file safety checks. Content that exceeds the automatic text-review limit is sent for manual review.


Image & file moderation

When you upload an image or document inside a receipt, Vid Receipts runs an automatic safety check before the file is stored or displayed to anyone.

How it works

  1. You pick a file to upload in the receipt editor.
  2. Before upload, the file is scanned for:
    • Format and size: only allowed file types up to 25 MB are accepted.
    • Content safety: images are analyzed by a multimodal AI vision model to detect harmful, graphic, or policy-violating visual content.
  3. If the file passes both checks, it is stored and embedded in your receipt.
  4. If the file fails either check, you'll see an error and the file is not uploaded.

What is checked

CheckWhat it catches
File typeImages: JPEG, PNG, WebP only (GIF is handled via the GIF picker, not direct upload). Documents: PDF only
File sizeMax 25 MB per file
Visual safetyNSFW content, graphic violence, dangerous content, harassment

The safety model

Image moderation uses a multimodal AI vision model tuned to block content at medium severity and above across four safety categories. This means:

  • Content that falls into medium, high, or extreme harm categories is always blocked.
  • When uncertain, the system errs on the side of safety, it fails closed, not open.

Your experience

  • Safe file: Upload completes normally. The image appears in your receipt.
  • Blocked file: You see a clear error message explaining the content was rejected. The file is not stored or shown to anyone.

If your content is flagged

Text receipt flagged:

If the AI rejects your receipt, it stays out of the public feed. A notification explains the current reason. You can edit and resubmit it. If an automatic review cannot finish, the receipt waits for manual review.

Image upload blocked:

If an image is blocked, the upload is canceled immediately. You can try:

  • Removing or replacing the flagged visual content
  • Using a different image that meets our content guidelines

We're building a formal appeal process. In the meantime, if you think your content was flagged by mistake, reach out to our support team and we'll take a look.


Tips for passing moderation

Writing great receipts is the best way to avoid moderation issues:

  1. Stay on topic: Make sure your receipt relates to the video
  2. Be specific: Vague or generic annotations are more likely to be flagged
  3. Cite sources: Back up claims with links or references when possible
  4. Keep it respectful: Disagree with ideas, not people
  5. Add value: Ask yourself: "Would this help someone watching this video?"

For images: upload screenshots, diagrams, or reference images relevant to the video. Avoid anything graphic, sensitive, or unrelated.


Notifications after approval

A comment notifies the receipt owner after both the comment and receipt are approved. Receipt notifications also wait for approval. Restoring the same receipt or comment does not send another notification.

Notifications use the posting persona. If content becomes unavailable or its author deletes their account before delivery, the queued notification is withheld. Email bursts keep the same timing: one immediate message, followed by a grouped message after activity settles. The grouped message links to your current notifications without quoting old content.

Your privacy

Text is sent to the moderation service when automatic review is needed. The service can use your prior moderation outcomes to decide whether further review is necessary. Receipt and comment notifications use the posting persona, and a result for an older edit cannot publish a notification about the current version.

For more details on how we handle your data, see Data & privacy.

Related

Last updated on