NEWS & INSIGHTS

For Writers (Articles, scripts, marketing copy)

Before drafting

• Write the angle + thesis in your own words before you prompt.
• List 3–5 “must-include” original elements (your anecdotes, unique POV, proprietary framework, interview notes).

During generation

• Use AI for variation, not voice (headlines, outlines, alternate phrasings), then re-author the final.
• Avoid “write in the style of…” prompts—instead ask for tone attributes (clear, punchy, authoritative).

Editing & originality

• Do a structural rewrite pass (reorder sections, replace examples, rewrite the intro + conclusion yourself).
• Run a “signature pass”: add original metaphors, examples, and transitions that reflect your voice.
• Quote/fact discipline: independently verify stats, dates, names, claims, and quotes.

Rights & disclosure

• Don’t paste confidential drafts, client materials, or unpublished manuscripts into prompts unless approved.
• Track sources used for research inspiration and make sure you’re not inheriting protected phrasing.

Documentation

• Save: outline → AI outputs → your rewrite drafts → final. Keep a short log of what you authored vs. what AI suggested.

For Designers (Brand, web, product, illustration, video)

Inputs & reference hygiene

• Only use reference assets you have rights to (stock licenses, client-owned, your own work).
• Avoid prompts that request a specific artist’s recognizable style; request visual attributes instead (color palette, lighting, composition).

Generation & iteration

• Treat AI as a rough comp tool: explore concepts fast, then rebuild/compose with your own decisions.
• Make the “final arrangement” human-led (layout, typography, hierarchy, spacing, color decisions).

Brand & consistency

• Lock brand tokens (palette, type, spacing rules) and ensure AI outputs don’t drift.
• If output resembles an existing brand/artist too closely, regenerate with stronger constraints or rebuild from scratch.

Rights & licensing

• Confirm downstream usage is permitted: commercial use, derivatives, exclusivity (especially for campaign/packaging).
• For video/audio elements (music, VO), verify licensing terms and model usage policies.

Documentation

• Save: moodboards + prompts + iterations + your composition files + export history. Note what is AI-generated vs. handcrafted.

For Developers (Code, docs, datasets, models)

Confidentiality & security

• Never paste secrets (API keys, tokens, private repo URLs, credentials) into prompts.
• Don’t share proprietary code blocks unless you’re using an approved environment and policy allows it.

Correctness & provenance

• Treat AI code as untrusted until verified:
  • run tests,
  • review for edge cases,
  • validate dependencies and licenses.
• Avoid copy-pasting large blocks without understanding them—create a quick “explain + threat model” pass.

License & IP checks

• Scan dependencies AI introduces (package licenses, copyleft risks).
• If generating from a well-known library pattern, double-check you’re not inheriting licensed snippets verbatim.

Quality gates

• Require: unit tests, linting, security checks (SAST), and peer review before merge.
• For generated docs: confirm that examples compile/run and align with real APIs.

Documentation

• Keep: prompt history for key decisions, code diffs, test results, and rationale notes—especially for customer-facing or patented work.

A Simple “Go / No-Go” Gate for Everyone

Go if: you can clearly explain your human decisions, inputs are clean, and the final feels meaningfully authored by you.
No-Go if: it imitates a specific creator too closely, relies on unverified claims, or uses unclear-rights inputs.