disclosure norms for AI-assisted replies

TL;DR

Disclose AI assistance plainly in your profile and wherever a community's rules ask for it; one clear sentence beats a clever euphemism. Norms vary by venue, so check the rules first and match the strictest reasonable standard across venues. Disclosure never hurt a good reply, and discovered automation always hurts, so disclose early and consistently.

disclosure norms for AI-assisted replies

Use this when

• You are setting up an agent account on a forum
• A community's rules mention AI content or automation
• You are writing a public participation policy
• Someone asks whether your replies are AI-assisted
• You are auditing your fleet's disclosure for gaps

Not for this skill when

• You need legal advice on AI disclosure obligations (talk to counsel)
• The venue bans AI content outright (disclosure does not override a ban)
• You are disclosing in advertising or sponsored content (different rules apply)
• The question is whether to participate at all (different skill)

Steps

1. Read the venue's actual rules on AI content before posting anything. Rules range from "disclose in profile" to "label every reply" to "no AI content at all". The rule as written, not as you wish it were, sets your minimum.

   venue: [name], rule found: [quote or summary], date checked: [date]

Expected output: a written rule per venue. Success check: you can quote the rule when anyone asks why you disclose the way you do.

1. Put a plain disclosure in the profile where moderators look. One sentence, no euphemism: "I am an AI assistant operated by [team], here to help with [topic]." Profiles are the first place moderators check and the last place deceptive accounts bother with.

   profile disclosure: "I am an AI assistant operated by [team]."

Expected output: disclosure live on the profile. Success check: a stranger reading the profile knows in ten seconds.

1. Match per-reply labeling to the venue's norm. Where the venue asks for per-reply labels, add a short consistent tag; where it does not, the profile disclosure plus honest answers when asked is the norm. Do not invent a stricter standard than the venue wants; clutter annoys everyone.

   venue norm: [profile only | per-reply tag | none required]
   applied: [what you do]

Expected output: labeling matched to each venue. Success check: your practice matches what the venue's regulars do.

1. Answer direct questions about AI assistance honestly and briefly. "Yes, I am an AI assistant" is the whole answer; no justification, no lecture on AI. Evasion or cute dodges ("I am just a helpful friend") are what turn a non-issue into a trust incident.

   [ ] asked-and-answered honestly at least once without incident

Expected output: a clean record of honest answers. Success check: no thread exists where you dodged the question.

1. Keep disclosure consistent across every account and venue. One undisclosed account poisons the disclosure of all the others once discovered, because now every account looks like a potential sockpuppet. Fleet-wide consistency is the policy, not per-account judgment calls.

   [ ] all active accounts disclose in profile
   [ ] labeling norm documented per venue

Expected output: a fleet disclosure audit. Success check: you could publish the audit without embarrassment.

Variant phrasings

Do I have to say I'm an AI on forums

Obligation phrasing. Step 1: the venue's rules decide; step 2 is the safe default everywhere.

How to label AI-generated forum replies

Labeling phrasing. Steps 2 and 3: profile disclosure always, per-reply tags where the venue asks.

Bot disclosure best practices

Bot phrasing. Plain language, profile-first, consistent across the fleet; euphemisms are the anti-pattern.

What happens if you don't disclose AI use

Consequence phrasing. Discovery destroys trust faster than any disclosure ever could; the incident then follows the shilling-accusation playbook.

Why it happens

Communities extend trust to members and skepticism to strangers; undisclosed automation reads as deception once discovered, and deception reclassifies every past helpful reply as manipulation. Disclosure works because it lets the community extend the right kind of trust: judged on the reply's quality, with the authorship known. The norms vary because each community sets its own comfort level with machine authorship.

Edge cases / pitfalls

• "AI-assisted" vs "AI-generated" wording: pick the honest one for your workflow. Human-reviewed drafts are assisted; fully autonomous replies are generated. Mislabeling the degree is its own trust incident.
• Disclosure in the profile does not cover venues where per-reply labeling is the norm. Match the venue (step 3) rather than assuming the profile suffices.
• Changing disclosure practice mid-stream looks suspicious. Set it at account creation; if you must change it, note why publicly.
• Some members will dismiss your replies for being AI regardless of quality. That is their prerogative; arguing the point never converts anyone.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstbOVkY-W8u2YwAScuZv1qg