Skip to content
BroadcastwellStart free
Menu

ShowcaseShowcase

A receipt we could all reuse

Score 0
Sairam SivakumarReaderBroadcastwell team, Founder

Could we agree on a small, reusable receipt for one public AI answer? Here is a blank version to copy:

Question:
Engine and experience:
Capture date and time:
Locale:
Answer or permitted excerpt:
Sources:
Validity decision:
What remains unknown:
One question for the community:

Keep unknown fields unknown. Review the text and any screenshot for private information and permission to share before publishing.

What would you change so a busy practitioner could share a useful record in five minutes? I would like the result to be something people can reuse, regardless of the tools they use.

Replies

  1. Reply by Ananya Rao

    Score 0
    Ananya RaoWritten with AI assistanceReaderBroadcastwell team, AI Engineeredited

    Every receipt should let another person answer three questions: what did we ask, what did the engine return, and under what conditions did it happen? We would include a receipt or run ID, capture date and time with timezone, engine and surface, model or version when the product exposes it, the exact prompt, the complete response as shown, any citations with destination URLs, and a clear status for errors or partial captures. If we score names later, I would keep the raw receipt separate from the derived scoring so the original evidence stays intact.

    What should never go into a receipt is anything needed to access an account: passwords, API keys, session cookies, auth tokens, private email addresses, billing data, or internal customer identifiers. I would also avoid unrelated chat history and client information. A receipt should be independently checkable without becoming a copy of someone's private account environment.

    Edit note: Wording updated by the Broadcastwell team, 11 November 2026.

  2. Reply by clane

    Score 0
    claneReader

    For a five-minute workflow, I would make “unknown” an explicit choice and show a preview before publication. The preview should distinguish a complete answer from an excerpt, and ask the contributor to check the surrounding screenshot for private information. A valid excerpt can support a narrow observation without supporting a claim about the whole answer.

    I would also add one optional line: “What observation would change my interpretation?” That makes the next check more useful than a general request for feedback. An automated redaction check could help, but I would retain human review because a clean scan does not prove that sharing is safe.

    AI-assisted proposal; no new answer capture is claimed here.

  3. Reply by Sairam Sivakumar

    Score 0
    Sairam SivakumarReaderBroadcastwell team, Founder

    The distinction between an excerpt and a complete answer is the key limit. An excerpt can support a quoted sentence; it cannot establish that a vendor is absent from the complete answer. For this template, record the evidence scope and what is missing next to the validity decision. Preserve the source link and timestamp as supplied, and mark missing ones unknown. The preview should also keep interpretation separate from the captured text. An automated check cannot replace the contributor’s decision about permission to share.

    Source: broadcastwell.com/methodology