AI Prompts for Investigative Journalists

Using AI Without Compromising Verification Standards

Investigative journalism demands rigor AI cannot replace—but AI can accelerate document analysis, source prep, and narrative structure when prompted precisely. The key is using AI for synthesis and scaffolding, never for fact generation.

1. Document Triage and Pattern Extraction

Prompt: "I have 40 pages of procurement contracts. List every instance where: (a) the vendor name changes but the address or principal is the same, (b) the contract value is just below a reporting threshold (e.g., $24,999 when the threshold is $25,000), (c) any date appears inconsistent with the contract timeline. Return findings as a numbered list with page and line references. Do not infer—only flag what is literally present in the text. [Paste document.]"

Why it works: The 'do not infer' instruction is critical for legal and editorial defensibility. You want a pattern-spotter, not a fabricator.

2. Source Interview Prep

Prompt: "I'm interviewing a former CFO of a regional hospital about Medicaid billing irregularities. Based on this background memo [paste], generate 12 questions organized as: 4 rapport-building questions about their career, 4 factual clarifiers about the documents I'll present, and 4 accountability questions I may need to push on. Flag which questions might cause the source to shut down."

Why it works: The 'flag shutdown risk' instruction gives the journalist tactical awareness during a live interview—something most reporters develop only through years of practice.

3. FOIA Request Draft

Prompt: "Draft a FOIA request to [agency name] seeking all communications between [official] and [company] from January 2022 to present. Include: email, text, calendar entries, and meeting notes. Use precise legal language that minimizes agency discretion to claim 'no responsive records.' Cite 5 U.S.C. § 552 and include an expedited processing request based on public interest. Keep under 400 words."

Why it works: FOIA language is technical. AI trained on legal text can produce stronger statutory citations and narrower request framing than most reporters write from scratch.

4. Narrative Structure for a Long-Form Piece

Prompt: "I have a 5,000-word investigative piece about [topic]. Here is my current draft outline [paste]. Identify: where the reader is most likely to disengage, whether the inciting anecdote is strong enough to carry the introduction, and whether my central finding appears too late. Suggest two alternative structural approaches with a one-paragraph rationale for each."

Why it works: Structural feedback is one of the highest-value editing tasks AI can perform. Unlike fact-checking, it doesn't require access to source material.

5. Legal Sensitivity Review

Prompt: "Review this draft paragraph for potential defamation exposure. Flag any statement of fact about a named individual that: (a) could be false, (b) is not sourced in the text, or (c) implies wrongdoing without documented evidence. Do not rewrite—only flag and explain the specific risk. [Paste paragraph.]"

Why it works: Legal review AI can't replace a media lawyer, but it can catch obvious exposure before the expensive consultation. The 'do not rewrite' instruction keeps the journalist in control.

The Non-Negotiable Rule

Never use AI to generate quotes, paraphrase sources, or fill gaps in documentation. Use it only where human pattern-matching is slower than the deadline.

Grade and improve your prompts free at http://143.198.136.81.nip.io:8802