Agents

Claim checker GPT that extracts factual claims from a draft and verifies each against sources

A custom GPT for writers, editors and comms teams: paste a draft and it lists every checkable claim, searches for supporting or contradicting sources with web search, and returns a verdict table with links, so you fix problems before publishing.

Agent in Research by AISkillHub Library

What it does

Drafts pick up unverified claims: a statistic from memory, a date slightly off, a quote attributed to the wrong person. This GPT pulls every checkable claim out of a draft and checks each one, showing its sources so an editor can make the call.

**What it does**

1. **Extracts claims:** numbers, dates, names and titles, quotes, historical facts, scientific or technical statements, comparisons and "first/only/largest" claims. Opinions and predictions are listed separately as not checkable. 2. **Prioritizes:** claims that would be most damaging if wrong (numbers, quotes, legal or health statements) are checked first. 3. **Checks** each claim with web search, preferring primary sources (official statistics, original studies, company filings, the original speech or publication). 4. **Returns a verdict** per claim: Supported, Partly supported, Contradicted, Unverified, or Outdated, with the source links and a short note on any mismatch. 5. **Suggests fixes:** corrected wording, a hedge, or "cut or source this".

**When to use it**

- Blog posts, newsletters, reports, speeches and press materials before publishing - Checking a document someone else wrote before you put your name on it - Updating an old article whose figures may be out of date

**What it will not do**

- It does not confirm a claim because one website repeats it. Unverified means no reliable source was found, and it says what it searched. - It does not check claims that depend on your internal data (your revenue, your customer counts). It lists them for an internal owner. - It is not legal review. Defamation, privacy and regulatory concerns are flagged for a human. - Web results can be wrong or incomplete; the editor must open the links. The GPT never presents its own memory as a source.

**Example row**

| # | Claim | Verdict | Source | Note | |---|---|---|---|---| | 3 | "The policy took effect in 2019" | Contradicted | (link to official notice) | Official notice gives 2020; 2019 is when it was announced |

How to use it

**Set up as a ChatGPT custom GPT** (requires a ChatGPT plan that lets you create GPTs).

1. In ChatGPT, open the GPTs area from the sidebar (Explore GPTs) and choose **Create**. 2. Switch to the **Configure** tab instead of using the chat-based builder. 3. Set **Name** to `Claim Checker` and write a one-line **Description**. 4. Open `assistants/draft-claim-checker/instructions.md`, copy everything, and paste it into the **Instructions** field. The file is written to fit inside the field's character limit. 5. Add **Conversation starters**: `Check the claims in this draft:`, `Only check the numbers in this:`, `Is this quote accurately attributed?`, `Which claims here need an internal source?` 6. Under **Knowledge**, upload (optional) your organization's approved fact sheet or boilerplate, so internal claims can be checked against it. The GPT treats these files as internal sources and labels them that way. 7. Under **Capabilities**: turn **Web search** on (required for this GPT to work). Image generation and Code Interpreter are not needed. 8. Test it in the preview pane on the right with a real example, then save and choose who can access it (only you, anyone with the link, or your workspace, depending on your plan).

Files

Save each file at the path shown, relative to your project root.

  • assistants/draft-claim-checker/instructions.md3.3 KB

More agents

See all
  • Agent· SecurityFree

    Security reviewer subagent for branch diffs

    A Claude Code subagent that reviews the changes on a branch for exploitable security issues (injection, broken access control, secrets, SSRF, path traversal, unsafe deserialization, weak crypto) and reports each with an attack scenario and fix.

    Coding agentsClaude

    ALAISkillHub Library

  • Agent· EngineeringFree

    TypeScript strict-mode migration subagent

    A Claude Code subagent that turns on TypeScript strict checks one flag at a time and fixes the resulting errors directory by directory with real types, without falling back to `any` or blanket `@ts-ignore`.

    Coding agentsClaude

    ALAISkillHub Library

Get AISkillHub updates in your inbox.

New features as they ship, AI live chat improvements, and the best new listings in the catalog.

Product updates, new features, and guides. Unsubscribe any time.

Claim checker GPT that extracts factual claims from a draft and verifies each against sources - AI agent - AISkillHub.dev