In short: Auto-Moderation rules label new posts for you based on conditions you choose, so your team only reviews what needs a human decision.
Understanding Auto-Moderation rules
Auto-Moderation rules take repetitive labelling off your moderators. Each rule combines conditions, such as an AI image signal, a keyword or an account detail, and applies a label to every new post that matches all of them.
A good set of rules does three things:
Protects genuine sellers: whitelisting rules label your own and approved accounts first.
Catches clear infringements: detection rules label obvious counterfeits, impersonation and copycats.
Shrinks the queue: clean-up rules set aside posts with no brand signal.
You'll find your rules in Settings → Moderation → Auto-moderation.
1. Available conditions
Conditions are the building blocks of a rule. Combine several to describe a very specific infringement pattern.
Image conditions
Condition | What it checks |
Image Features | The brand features Zeal's AI found in the image, such as your logo or a signature pattern |
Image Feature Type | How the AI classed what it found: Highly Suspicious, Obvious Counterfeit or Logo |
Logo Detected | Whether a logo appears in the image |
Image Label | The label the image has been given |
Image Label And Status | The image's label together with where it is in the workflow, for example Counterfeit and Validated |
Post conditions
Condition | What it checks |
Post Description and Title | Words in the post's title or description |
Post Title | Words in the title only |
Post Listed Brand | The brand named in the post |
No Listed Brand | Whether the post names no brand at all |
Post Has Obfuscated Brand Name | Whether the brand name is disguised or misspelled, for example "s0lene" |
Post Abnormal Price | Whether the price falls outside the expected range |
Price | The post's price |
Post Product Condition | New, used, refurbished and so on |
Post Stock Count | How many items the seller has available |
Post Label | The label the post has been given |
Post Is Checked | Whether the post has been reviewed |
Website conditions
Condition | What it checks |
Website Name | The website or platform the post is on |
Website Category | The type of website, such as Marketplace or Social Media. This is the same data point as the Channel filter |
Website Is Tracked | Whether the website is set up in Settings under Crawling Domain |
Account conditions
Condition | What it checks |
Account Name | The seller's account name |
Account Country | The seller's registered country |
Account Follower Count | How many followers the account has |
Account Label | The label the account has been given, such as Authorized Seller |
Account Tag | Tags added to the account |
Account Is Checked Or Validated | Whether the account has been reviewed or validated |
Operators
Each condition uses one of three kinds of operator:
Any In / None In: for lists, such as keywords, image features or labels. None In excludes something, for example posts with no logo.
Is True / Is False: for yes/no signals, such as Post Abnormal Price. Choosing the operator is enough; there's nothing to tick.
Greater than / Less than: for numbers, such as Price or Post Stock Count.
Learn more: AI Image Feature Detection · Channel Filter
2. Creating a rule
Build a rule whenever your team keeps applying the same label to the same kind of post.
Go to Settings → Moderation → Auto-moderation and click the + icon.
Enter a Rule Name that describes what the rule catches, and choose the Label to apply when it matches. Labels can be any you use in your Program, such as Counterfeit or Legitimate.
Under Conditions, click +, choose a condition, pick an operator and add the values. Repeat for each condition. A post must match all of them for the rule to apply.
Open the Preview tab to see which posts the rule would catch, then click Save.
Example: hidden brand name at a suspicious price
Post Has Obfuscated Brand Name → Is True
Post Abnormal Price → Is True
Account Label → None In → ["Official Store", "Authorized Seller", "Legitimate"]
Label: Counterfeit
Tip: Lean on Zeal's AI signals, such as Image Feature Type and Post Abnormal Price, rather than long keyword lists. They stay accurate when sellers change their wording.
Learn more: Auto-Moderation Rule Examples · Rule Suggestions
3. Setting rule priority
Priority decides the final label when a post matches more than one rule: the rule highest in your list applies.
Drag rules up or down the list to change their order. As a rule of thumb:
Whitelisting rules at the top, such as your official accounts labelled Official Store.
Detection rules next: impersonation first, then counterfeits, then copycats and grey market.
Clean-up rules at the bottom, such as posts with no brand signal labelled Not Relevant.
Important: If a detection rule sits above your whitelisting rules, it can label your own authorised sellers as infringers.
4. Activating a rule
A saved rule does nothing until you switch it on. Turn on the rule's toggle to make it Active.
Note: Active rules apply only to new, incoming posts that haven't been labelled yet. Posts already in Zeal before the rule was activated aren't moderated by it.
5. Measuring rule accuracy
The Rule Precision Score shows how often a rule's labels turn out to be right, so you know which rules to trust and which to tune.
Finding the score
Go to Settings → Moderation → Auto-moderation, open a rule with the pencil icon and select the Preview tab.
How it's calculated
Zeal compares the rule's labels with the labels your team confirmed during Review:
True positive: the rule labelled a post Counterfeit and the Quality Analyst kept that label. The rule was right.
False positive: the rule labelled a post Counterfeit and the Quality Analyst changed the label. The rule was wrong.
For example, if a rule labels 100 posts and 90 are confirmed, its Rule Precision Score is 90%.
Reading the score
First check how many reviewed posts the score is based on:
500 or more: high confidence. Use the score to make firm decisions about the rule.
Fewer than 500: a directional signal. The score may still move, so use it to spot problems rather than to delete a rule.
Then use these thresholds:
Score | What it means | What to do |
90–100% | High precision | Safe to rely on for automatic labelling |
70–89% | Moderate: about 2 or 3 in every 10 labels are wrong | Useful, but add a condition to make it more specific |
Below 70% | Low precision | Tune it straight away; it's creating extra work for your team |
Learn more: Understanding the Zeal Workflow
Quick reference
If you want to… | Look at… |
See what a rule can check | 1. Available conditions |
Build a new rule | 2. Creating a rule |
Stop a rule labelling genuine sellers | 3. Setting rule priority |
Know which posts a rule will label | 4. Activating a rule (new posts only) |
Check whether a rule is accurate | 5. Measuring rule accuracy |
Copy a ready-made rule |