Reporting a bad answer is not the same as appealing a ban

Start here

When an assistant produces something harmful, wrong in a way that could mislead someone, or shares content that should not have been shared, most people do not know there is anywhere specific to send that, beyond a thumbs-down button. There usually is a specific path, described in each vendor's own help pages as retrieved on 16 September 2026, and it is worth knowing before the moment it is needed.

What the documents say

Anthropic's support article on how to report, block, and remove content describes several separate channels depending on what happened: a direct safety-team email address for reports needing enough detail 'for replication,' an in-product report button for shared Claude content, and a standalone reporting form for output that 'violates Anthropic's usage policies or local laws.' The same article states a consequence for misuse of the system itself: Anthropic 'reserves the right to suspend users who repeatedly provide illegal content' and may stop processing notices from anyone who 'frequently submit notices which are manifestly unfounded.' Google's Gemini Apps guide describes a comparable but distinctly labeled path: selecting 'More' then 'Report legal issue' under a response, or a separate general feedback flow for lower-stakes complaints about accuracy or quality. It states that human reviewers 'assess if Gemini Apps' responses were low-quality, inaccurate, or harmful' and that reviewed content is retained for up to three years, a specific retention figure neither company frames as changeable by request.

Check this

The useful distinction to check is which channel matches the actual problem, since both vendors separate 'this output was low quality' from 'this violates policy or law,' and route them differently. A reader facing genuinely harmful output should look for the policy-violation or legal-issue path specifically, rather than a general feedback thumbs-down, since the documents describe that channel as the one tied to formal review and, per Google's page, longer retention for investigation. Confirm what a report actually claims: Anthropic's language about suspending repeat providers of illegal content is a two-way consequence worth understanding before filing a report that is itself unfounded.

What holds and what fails

Both documented systems hold as a real channel distinct from ordinary feedback, backed by a stated human review step rather than an automated dismissal. Neither document states a guaranteed response time or outcome for an individual report, and neither describes an appeals path for a user's own suspended account within the same article, meaning a reader banned or restricted may need to locate a separate process not covered by these two sources. This is a gap worth naming plainly rather than assuming a report or an appeal will be resolved quickly.

• Locate your assistant's specific policy-violation or legal-issue reporting path before you need it.
• Include enough detail to allow reproduction, as Anthropic's article specifically requests.
• Understand that repeatedly filing unfounded reports carries its own stated consequence.

A reporting path only helps if it is used correctly and found in advance; both companies describe real review processes, but neither promises a fast or certain result for any single report.