How answers work

Ask a small question

As People Say handles short Italian and English language uncertainties. You choose what you have and what you want to learn, and the composer maps that intent to one of four formats: checking a sentence, comparing two choices, asking how to say something, or asking what something means.

Situation details such as audience, channel, and register help speakers judge what fits. Remove identifying details from quoted messages. Before submitting, you can keep a question private or hide your name. A monolingual question uses one request; a bilingual question uses two because it requires scarcer attention.

Qualified people answer independently

The Help flow routes one question at a time to members whose stated language abilities meet the task. You cannot choose questions from an answering queue, answer your own question, or contribute twice to the same question.

For a sentence check or choice, speakers answer independently and cannot see earlier judgments. For a bilingual question, one qualified speaker writes a candidate and other qualified bilingual speakers review it. Reviewers can support it or offer a different wording; an author cannot support their own candidate.

Every contribution must be the person's own work. The service does not generate, translate, or summarize user content with AI.

The system gathers bounded evidence

A result is produced when the configured settlement rules have enough evidence or the question reaches its contribution limit or expiry. More answers are requested when early judgments disagree, but every format has a hard cap so a question cannot consume unlimited attention.

Resolved does not mean objectively certain. Results say how strong the evidence was: clear consensus, consensus, leaning, mixed, or insufficient. A mixed result is useful evidence that real speakers disagree. An insufficient result means the service did not gather enough evidence to say.

Open questions never reveal partial votes or emerging outcomes. This protects independent judgment. Once resolved, the result freezes the evidence used, shows the leading answer or interpretation first, and then shows distributions, supported alternatives, edits, and optional speaker notes where relevant.

Asking stays reciprocal

Your first question is available without helping first. After that, helping learners earns asking capacity. Quick judgments, bilingual reviews, and seed answers earn different internal attention units according to their expected effort. The interface expresses the balance as requests, not points.

Helping is never limited by your balance, and stored capacity has a small cap. Skipping costs and earns nothing. If there is nothing you are eligible to help with, a limited grace ask may become available. The composer shows the price and current wait estimate before submission.

Questions that expire with insufficient evidence and questions withdrawn before any relevant contribution receive the specified refund. A question closed because qualified speakers say it cannot be answered as asked normally does not.

Privacy and accountability

Non-private questions and resolved answers are readable only by signed-in members. Open Explore pages do not show contribution evidence, and private questions never enter Explore. Anonymous mode hides the requester's identity but not the question itself.

Members can report visible content. Moderation is handled by people, recorded, explained to the affected member, and appealable. Read the Rules, Privacy notice, and Terms for the full draft policies.