How AmaQnA works

1. 1

   Find a host

   Browse the discovery feed or open a host page directly (for example /h/their-slug).
2. 2

   Write your question

   Be specific. One clear question gets a better answer than three stacked together. Add your email if you want a ping when it's answered — it's optional and never shown publicly.
3. 3

   Optionally boost it

   Boosting moves your question to the top of the host's queue. It doesn't guarantee an answer, but it signals priority.
4. 4

   Read the answer with its label

   Every answer carries one of four labels so you know exactly how much the host stands behind it. See the four labels.

A host page won't accept questions until you've (a) consented and recorded a voice sample and (b) approved at least a small library of source material. This is enforced in the database, not just the UI — no guessing in your name.

1. 1

   Claim your slug

   In Dashboard → Settings, pick your public URL and add a bio. This is what askers see at /h/your-slug.
2. 2

   Record or upload your voice

   Under the Voice tab, record 2–5 minutes of natural speech, or upload a clean audio file (up to ~10 min / 25 MB). See voice tips.
3. 3

   Add knowledge sources

   Under Knowledge, paste text, upload PDFs/DOCX, drop in URLs, or connect your website and social profiles. Aim for at least eight solid atoms to unlock the public page. See source tips.
4. 4

   Review drafts in your inbox

   Every incoming question is drafted by multiple AI models, grounded in your sources, then queued for your review. Edit, approve, or escalate. See the answer inbox.

• Length: 2–5 minutes of natural speech is the sweet spot. You can go up to 10 minutes.
• Format: MP3, WAV, M4A, OGG, FLAC, or WebM. Max 25 MB.
• Quality: quiet room, one voice, no music or overlap. A phone mic held ~15 cm from your mouth is fine.
• Content: read or speak how you actually talk — an interview, a talk excerpt, or a candid explanation of your work all work well.
• Consent & warranty: you'll confirm you own the rights to the voice you upload. Voice replies always open with an audible AI disclosure and ship with a C2PA-shaped provenance manifest in the audio.

Your knowledge twin can only answer from material you've approved. More depth = better, more grounded answers.

• Paste text — the fastest way to seed a topic (essays, talk transcripts, notes).
• Upload files — PDF or DOCX up to 15 MB, plain text up to 2 MB. Drag-and-drop multiple at once.
• Crawl a URL — a blog post, article, or public page.
• Website & social profiles — connect your site, X, LinkedIn, Instagram, YouTube, or Spotify and we'll crawl what's public.

Sources are chunked, embedded, and stay private to your account. They are never used to train third-party models.

1. An asker submits a question on your public page.
2. We search your knowledge atoms for the most relevant passages.
3. Multiple AI models draft an answer independently. We pick the strongest and score cross-model agreement.
4. The draft lands in your inbox with citations, an agreement score, and a preliminary label.
5. You edit, approve (labelled Personally approved), escalate for a personal reply, or re-draft.
6. Approved answers can be synthesised in your cloned voice with an audible AI disclosure.

The queue updates in realtime and can send a browser notification when a new question arrives.

Asking is free. Boosts let askers pay to jump the queue; if a host can't answer within the promised window the boost is automatically refunded.

Hosts start on a free plan. Paid plans unlock higher volume, custom domains, the embed widget, and API access. See /pricing for the current tiers, and /for-hosts for revenue-share details.

A host can issue an API key so a publisher, community, or product can fetch provenance-labeled answers as JSON. Each call consumes one request from the key's monthly quota.

Query params: limit (max 100), cursor for pagination.

{
  "host": { "slug": "jane-doe", "display_name": "Jane Doe" },
  "items": [
    {
      "id": "a1b2c3...",
      "question": "How do you approach product-market fit?",
      "answer": "I look for...",
      "audio_url": "https://.../answer.mp3",
      "permalink": "https://amaqna.com/h/jane-doe/q/123",
      "updated_at": "2026-07-02T09:30:00Z"
    }
  ],
  "next_cursor": "2026-07-02T09:00:00Z"
}

Responses include x-ratelimit-limit and x-ratelimit-remaining headers. The widget uses the same endpoint.

• Voice cloning requires explicit consent and a rights warranty.
• Every synthesised reply opens with an audible AI disclosure in the same voice, plus a C2PA-shaped provenance manifest.
• Answers are grounded in the host's own sources — we won't invent quotes.
• Questions containing PII are automatically scrubbed before they reach the host.
• Hosts can delete their profile, sources, voice model, and answers at any time from the dashboard.
• We do not sell content to third parties for training.

Full detail lives in Privacy, Terms, and Acceptable use.