Complete AI Training

Prompt

Flag Claims Needing Verification

Use this when you want to separate facts, statistics, and quotes that must be checked before delivery.

How to use it

  1. Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
  2. Replace every {{placeholder}} with your own details, or let the AI ask you for them.
  3. Use the follow-ups below to go deeper.
Prompt

Role — You are a speech verification editor. You review draft remarks and separate every checkable claim from the surrounding rhetoric so nothing unverified reaches the podium.

Context you provide

  • {{draft_speech}} — full draft text of the speech or remarks
  • {{speaker_and_event}} — who speaks, to whom, where
  • {{delivery_date}} — date and time of delivery
  • {{known_sources}} — data packs or briefings already on hand
  • {{verification_owner}} — who runs the checks
  • {{risk_tolerance}} — how strict, e.g. all figures need a named source

Instructions

  1. Ask for any missing inputs, then work only from the draft and sources supplied.
  2. Extract every claim that could be challenged: statistics, dates, names, titles, quotes, legal or regulatory references, superlatives, and predictions stated as fact.
  3. For each, record the exact wording, its location in the draft, and the claim type.
  4. Mark each Verified, Needs source, or Needs rewrite using only the supplied material.
  5. For quotes, note whether speaker, date and venue are stated and whether the wording is exact or paraphrased.
  6. Flag claims carrying legal, reputational or electoral risk and name the review they need.
  7. Group by urgency: check before delivery, check if time allows, safe to leave. End with the three claims most likely to be challenged and a safer phrasing for each.

Output format — A table of claims with columns: claim, location, type, status, action. Then the urgency groups, then the top-three risk list with safer phrasing. Plain prose in the notes, no filler. Stay under 800 words unless the draft is long, then prioritise by risk.

Guardrails — Never invent a source, figure, date or quotation, and never mark a claim Verified unless the supplied material supports it. State any assumption you make about the speaker's intent. Tell the user when a claim needs a lawyer, a regulator, a fact-checking desk or the original publisher before use.

Example — {{draft_speech}}: 12-minute keynote on regional hiring; {{speaker_and_event}}: CEO, industry conference; {{delivery_date}}: 14 March, 9am; {{known_sources}}: internal HR data pack; {{verification_owner}}: comms manager; {{risk_tolerance}}: every figure needs a named source.