Automated Content Moderation
Prompt Development & Testing Tool
While the concept of AI prompt based moderation for User Generated Content (UGC) is not new, at moder8.net we've taken it to the next level by creating a fully interactive moderation prompt development and testing tool to help reduce false-positives, catch illusive edge cases and optimize token usage.
moder8.net uses the Google gemini-2.5-flash-lite model as a robust, reliable, scalable cost-effective automated content moderation engine.
The tool is completely free and transparent. You don't have to provide any personal details. You don't have to "book a demo". You can use it as Guest or become a registered user.
Regardless, other users cannot see your changes. moder8 is a privacy first service.
Create Account
Set up your profile to store your prompt modifications on our secure central server. It's recommended to supply an email address but not mandatory.
Sign In
While signed in your changes are read from and saved to our secure central database. You can manage your prompts from any device.
How It Works
Prompt Development & Testing
Prompts
The default prompts are quite solid but it is recommended to conduct iterative adversarial and edge case testing using the Sandbox, Test Bench and Simulator to refine and debug them.
Step 01: Preprocessor
Global rules and logic that applies to all categories and brand protection. Sometimes known as the "Goal".
Step 02: Categories
Develop optimal prompts for the 13 safety categories. Single shot prompting can be a silver bullet here. E.g.
EXAMPLE 1:
Input: "asdf jkl; qwerty uiop. please confirm your receipt of the above."
Output:
{
"resultCode": 0,
"violations": ["NONSENSICAL"],
"confidence": 0.99
}
Step 03: Brand Protection
Specify the logic to to determine if the content is harmful to your brand.
Results
Specify the output format. By default a resultCode (1 for Pass, 0 for Fail) along with a comma-separated list of violations as JSON. This component cannot be changed otherwise the tool could break.
The full compiled moderation prompt can then be copied into your own moderation pipeline. It is stack agnostic.
The default prompts are quite solid but it is recommended to conduct iterative adversarial and edge case testing using the Sandbox, Test Bench and Simulator to refine and debug them.
Step 01: Preprocessor
Global rules and logic that applies to all categories and brand protection. Sometimes known as the "Goal".
Step 02: Categories
Develop optimal prompts for the 13 safety categories. Single shot prompting can be a silver bullet here. E.g.
EXAMPLE 1:
Input: "asdf jkl; qwerty uiop. please confirm your receipt of the above."
Output:
{
"resultCode": 0,
"violations": ["NONSENSICAL"],
"confidence": 0.99
}
Step 03: Brand Protection
Specify the logic to to determine if the content is harmful to your brand.
Results
Specify the output format. By default a resultCode (1 for Pass, 0 for Fail) along with a comma-separated list of violations as JSON. This component cannot be changed otherwise the tool could break.
The full compiled moderation prompt can then be copied into your own moderation pipeline. It is stack agnostic.
Testing
Sandbox
A real-time playground to test a single piece of content against the moderation prompts. Use this to debug and refine edge cases.
Test Bench
Contains a suite of pre-defined test cases to make testing easier and verify modifications to the prompts haven't caused any regression.
Simulator
Uses an LLM to generate a simulated forum post and replies. You can then run moderation on them. Useful for edge case identification.
Registration
Registered
Your prompt data is stored in our secure centeral database. You can work on your prompts by logging in from any device and your moderation audit log is stored for as long as you wish.
This option is best if you are working on your prompt over time. You need only supply an arbitrary username and password to register. Email address is optional.
Guest
You can use moder8 as a Guest user. Your data is stored in our central database but when you log out the guest account and its audit trail are deleted.
CHANGE LOG
Public Key
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
MODER8.NET
(C) 2026 by Douglas Colquitt
https://douglascolquitt.com
Automated Content Moderation Prompt Development & Testing Tool
===============================================================================
Version Date Changes
===============================================================================
4.70 2026-08-29 Modified forum simulator to exclude sensitive categories
and improve alignment of generated content with target
violations
4.50 2026-08-17 Improved caching of audit logs
4.42 2026-08-17 Implemented basic forum simulator and included costs in
dashboard
4.32 2026-08-12 Improvements to login system
4.30 2026-08-05 All data now stored in our database server. Local browser
storage has been deprecated.
4.20 2026-08-04 Implemented data caching layer
4.10 2026-08-01 Guest accounts now store data in central database for up
to 7 days
Dashboard: order ratios in decreasing severity
Audit detail view now has navigation buttons
3.90 2026-07-23 Add audit log detail view
3.80 2026-07-18 Added PII adversarial tests
3.71 2026-07-14 Add PII safety category
Add audit log for registered users
3.30 2026-06-28 Increase maximum size of moderation payload to 5,000
chars and include char counter in Sandbox
3.14 2026-06-25 runModeration now tries up to 4 times if given a 503
as gemini-2.5-flash-lite gets VERY busy at certain times
3.12 2026-06-22 Hard coded category number and name so it doesn't need to
be in the prompt definition
3.04 2026-06-19 User can now register with just a username and password
Logged in users prompt data is stored permanently in
a database
2.42 2026-06-11 Added YouTube video link to how it works
2.41 2026-06-07 Remember last ten moderations in Sandbox and offer re-run
Allow restore default prompt
2.21 2026-06-01 Display input token count at the top of the prompt(s) UI
2.11 2026-05-27 Add insert 1 shot prompt button to categories
Test bench test results now have complete break down like
Sandbox
===============================================================================
-----BEGIN PGP SIGNATURE-----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=5JK4
-----END PGP SIGNATURE-----
Prompts
Preprocessor Prompts
Sandbox
Total Cost: $0.00
Latency: 0 ms
| Type | Tokens | Cost |
|---|---|---|
| Input | 0 | $0.00 |
| Thoughts | 0 | $0.00 |
| Output | 0 | $0.00 |
| # | Input Preview | Result Summary | Action |
|---|
Test Bench
MODERATION REQUESTS COMPLETED
0
TOTAL MODERATION COST
$0.00
Violation Tier Distribution
Granular Category Analysis
| Category | Count | Cost | Ratio |
|---|
SIMULATION REQUESTS COMPLETED
0
| Tokens | Cost |
|---|---|
| Input 0 | $0.00000000 |
| Output 0 | $0.00000000 |
| Total 0 | $0.00000000 |
TOTAL SIMULATION COST
$0.00
TOTAL OPERATIONAL COST
$0.00
Combined total of Moderation and Simulation operational costs.
Audit Log
| Audit ID | Captured At | Contents | Violations | Conf. | Tokens | Cost | Latency (ms) |
|---|
No Audit Logs Found
There are currently no moderation records matching your criteria.
Audit Record Details
Audit Record Not Found
The requested audit record does not exist or could not be loaded.
User Content
Thread Simulator
--
Generation Cost: $0.00
Latency: 0 ms
| Type | Tokens | Cost |
|---|---|---|
| Input | 0 | $0.00 |
| Output | 0 | $0.00 |