Bot Transparency Notices: How to Tell Members When Automation Acts

Explain what the bot can do, not what the team hopes it will do

A transparency notice is useful only when it describes the system members encounter. Saying that a community uses AI to keep discussions safe is too broad to explain whether a post can disappear automatically, whether an account can be restricted, or whether a moderator sees a warning before acting.

Start with the tool's actual role. It might identify likely spam, filter repeated links, flag content for moderator review, apply a temporary restriction, or remove a message that matches a narrow rule. These are different levels of authority and should be described separately.

For example, a community might use an automated filter to hold posts containing suspicious links while a volunteer reviews them. Another community might automatically remove messages containing a known spam pattern. Members should not receive identical notices for these two arrangements.

Describe the practical effect in ordinary language. If a message may be held before appearing publicly, say so. If an automated system can remove a message without an immediate human review, say that too. A notice that hides consequential details behind the word moderation is unlikely to help when a member's post disappears.

Separate automated flags from automated enforcement

A flag is a signal. It tells moderators that a message may deserve attention. Enforcement is an action that changes what a member can post, see, or do. Confusing the two creates false expectations and makes appeals harder to resolve.

A clear notice identifies each stage. For example: Automated filters may flag posts containing suspicious links or repeated promotional language. Flagged posts may be held for review. Moderators decide whether the content violates community rules, except for the specific automated filters described as immediate removals.

If a bot can take direct action, identify the scope. Members should know whether that action is limited to a particular category of content, whether it is temporary, and whether a human can review it. Avoid saying that all automated actions are reviewed by people if the team does not have the capacity to review every event.

This is also where a community rules template can prevent disputes. Put the general enforcement principle in the rules, then provide the technical explanation in a separate bot notice. The two documents should use the same terminology and describe the same decision-making authority.

Write a notice members can understand

A useful transparency notice should answer five practical questions: What tool is used? What kind of content does it examine? What actions can it take? When do moderators become involved? How can a member ask for review? The answers can be short, provided they are accurate.

Avoid claims such as the bot detects every violation, the system is unbiased, or false positives never happen. Automated classifiers can misunderstand jokes, quotations, specialized terminology, and context. Spam filters can also catch legitimate links or repetitive posts from new members. Explain the limitations without turning the notice into a technical manual.

A basic notice could read as follows: This community uses automated moderation filters to identify likely spam, suspicious links, and selected rule violations. Some posts may be held or removed automatically. Other flagged posts are sent to moderators for review. If you believe an automated action was incorrect, contact the moderation team through the appeal process described below. Moderators make decisions about disputed cases.

Change that wording if the actual configuration differs. If the bot only generates alerts, do not imply that it removes content. If automatic removals are not routinely reviewed, do not promise that every removal receives human attention.

Make the appeal route visible and usable

An appeal process is part of transparency because members need a way to correct mistakes. A notice that admits automated filtering but gives no route to challenge it describes a system without explaining how its decisions can be questioned.

Give members a specific contact method, the information they should include, and a reasonable expectation about the next step. Ask for the affected post or account, the action they are appealing, and a short explanation of why they believe the filter was wrong. Do not require members to reconstruct internal bot logs that they cannot access.

For example, a community could ask members to submit one appeal per affected post and state that a moderator will review whether the content violates the relevant rule. If the team cannot guarantee a response time, avoid promising one. A realistic notice can explain that urgent account restrictions receive priority and ordinary content appeals are handled as volunteer availability permits.

The person reviewing an appeal should be able to consider context beyond the original automated signal. If the original reviewer is unavailable, a second moderator can take the case. A bot should not be the final authority on whether its own decision was correct.

Explain data handling without making unsupported promises

Members may reasonably ask what information an automated moderation tool processes. A notice should identify the relevant categories, such as post text, links, account activity, or information made available through the platform's moderation interface. The exact description depends on the tool and its configuration.

Do not promise that no data is stored, shared, or retained unless the team has confirmed that those statements accurately describe the service and its settings. Some tools maintain logs, retain moderation events, or send content to an external service for processing. Product behavior and applicable requirements can vary by provider and region.

Keep the explanation proportionate. Members usually need to understand what content may be processed, the moderation purpose, who can access resulting records, and where to direct questions. They do not need a list of every technical setting. The team should consult the relevant provider documentation and applicable platform requirements before publishing specific data-handling claims.

Use a consistent notice across channels

A subreddit, forum, and chat server may present moderation information differently, but the underlying policy should remain consistent. A subreddit can place a short notice in its rules or pinned information. A forum can include it in the moderation policy. A Discord moderation setup can display a concise version in its rules or onboarding channel, with a longer explanation available in the server's policy information.

Assign one moderator to compare the public notice with the active configuration whenever the bot's permissions or behavior change. A notice that accurately described last month's setup can become misleading after an administrator enables automatic removals or adds an external processing service.

The next step is to list every automated action currently enabled, identify which ones require human review, and replace broad promises with precise descriptions. Publish the revised notice alongside the appeal route before expanding the bot's permissions.

Questions readers ask