Moderation allows teams to review, approve and control AI generated output before it is synchronized or published.
AI Workflows are designed to automate catalog operations while still allowing human oversight where needed.
Moderation helps maintain:
content quality
brand consistency
SEO standards
marketplace compliance
catalog accuracy
before AI generated content becomes live.
Why moderation matters
AI can automate large parts of catalog management, but not every workflow should publish content automatically.
Some workflows may require review because they affect:
customer facing content
SEO visibility
marketplace listings
translations
product categorization
Moderation creates a controlled approval layer inside the workflow pipeline.
How moderation works
When moderation is enabled:
AI generates output
the result enters a moderation state
editors or moderators review the output
approved results continue through the workflow
synchronized content updates the product data
Products may temporarily pause inside the workflow until moderation is completed.
What can be moderated?
Moderation can be applied to many workflow actions.
Examples:
Attribute Extraction
Content Enrichment
Translation
Category Mapping
Validation workflows
Quality scoring workflows
Any AI generated output can potentially be reviewed before synchronization.
Moderation roles
Depending on your workflow configuration, different users may participate in moderation.
Examples:
editors
moderators
workflow managers
Permissions and review responsibilities may vary depending on your setup.
Editor review
Editors typically review generated output for:
quality
formatting
consistency
correctness
Editors may:
approve output
decline output
review reasoning
inspect confidence scores
Moderator approval
Moderators may provide additional approval before synchronization.
This is especially useful for:
customer facing content
SEO critical pages
marketplace exports
high visibility product data
Moderator approval creates an additional quality control layer.
AI reasoning
Certain workflow actions display AI reasoning during moderation.
Reasoning explains why the AI generated specific output.
Example:
"The product description references salmon based cat food for adult cats, therefore Flavor was assigned to Salmon and Lifecycle to Adult."
Reasoning improves transparency and helps moderators evaluate AI decisions more effectively.
Confidence scoring
Moderation interfaces may also display confidence scores.
Confidence scores indicate how certain the AI is about the generated result.
Higher confidence usually means:
stronger source data
clearer context
more reliable output
Lower confidence may indicate:
ambiguous descriptions
missing information
uncertain classifications
Confidence scores help prioritize manual review.
Reviewing generated output
Moderation screens may allow reviewers to:
inspect generated values
compare source data
review AI reasoning
approve output
decline output
synchronize accepted changes
This gives teams detailed control over workflow quality.
Declining output
If generated content is incorrect or low quality, moderators can decline the output.
Common reasons include:
incorrect extraction
poor translations
formatting problems
SEO issues
invalid categorization
Declined outputs are prevented from synchronizing.
Synchronization after approval
Once content is approved:
synchronization becomes available
product data updates automatically
workflow execution continues
Approved content can then be pushed back into connected systems such as:
Magento
ecommerce platforms
marketplace integrations
Moderation queues
Large workflows may generate moderation queues.
Examples:
pending approvals
products awaiting review
synchronization waiting states
Moderation queues should be monitored regularly to maintain workflow throughput.
High moderation workflows
Some workflows benefit from stricter moderation.
Examples:
SEO content generation
marketplace descriptions
multilingual translations
homepage content
customer facing copy
These workflows often require brand and quality control.
Low moderation workflows
Other workflows may require minimal moderation.
Examples:
technical attribute extraction
internal validation
automated scoring
structured metadata generation
These workflows can often operate with lighter approval requirements.
Balancing automation and control
AI Workflows are designed to support different moderation strategies.
Examples:
fully automated pipelines
partially moderated workflows
fully reviewed approval pipelines
The right approach depends on:
catalog size
risk level
content visibility
operational capacity
Best practices for moderation
Moderate high visibility content
Customer facing content should usually receive stronger review processes.
Examples:
SEO pages
marketplace exports
shopping descriptions
translations
Use confidence scores strategically
Focus moderation effort on:
low confidence outputs
ambiguous products
unusual AI behavior
This improves moderation efficiency.
Optimize prompts over time
Frequent moderation corrections often indicate:
weak prompts
unclear workflow logic
insufficient source data
Improving prompts reduces moderation workload.
Separate workflows by moderation needs
Different workflows often require different approval strategies.
Examples:
automated attribute extraction workflow
moderated SEO workflow
lightly moderated translation workflow
This improves operational efficiency.
Example moderation flow
Example:
A webshop generates German Google Shopping descriptions.
Workflow:
Trigger selects products missing German content
Content Enrichment generates shopping descriptions
Translation converts content to German
Moderators review:
grammar
SEO quality
terminology
Approved content synchronizes to Magento
This creates a scalable but controlled localization pipeline.
Why moderation is important
Moderation allows businesses to scale AI automation safely.
Instead of manually creating all product content, teams can:
automate repetitive tasks
maintain quality standards
control synchronization
review uncertain outputs
continuously improve AI workflows
This creates a balance between automation speed and operational control.