1 · Topic selection

We only write topics that match one of four criteria:

• Real user pain: high search demand in our target language, scarce existing tutorials, recurrent in community Q&A
• YMYL defense: directly tied to asset safety, one mistake costs a lot. Example: 7 phishing patterns, Permit signature recognition
• Comparative tools with hands-on: alternatives differ significantly, beginners have no trusted comparison source
• Mechanism explained against AI-summary oversimplification: concepts that ChatGPT explains too simply

2 · Sourcing — three-tier hierarchy

1. Primary: project docs, EIPs / standards, whitepapers, on-chain explorers, DefiLlama, Dune, Chainalysis reports
2. Independent third-party audits / security firms: SlowMist, PeckShield, CertiK, Halborn, Trail of Bits, TRM Labs, Messari
3. Editorial first-party hands-on: every "editorial hands-on" block has date + time + device + numbers + verifiable tx hashes (archived)

3 · Review workflow · 3+1+1

1. Topic vote: 3 editors, single veto blocks
2. First draft: one author owns
3. Hands-on verification: a different editor reproduces the operation
4. Cross-review: the other two record all suggestions
5. External proofread: outside-editor reader-perspective pass
6. Final author sign-off; publish; archive who reviewed when

4 · AI tool boundaries

Editors use LLMs (ChatGPT, Claude, Gemini) for:

• ✓ Brainstorming article structure / outline discussion
• ✓ Grammar checks / sentence rewriting
• ✓ Cross-language draft translation (human-edited afterward)
• ✓ Source discovery (always human-verified)

Editors do not use AI to:

• ✗ Generate published prose directly (humans write, AI doesn't)
• ✗ Generate "facts" (LLMs confidently make up numbers; every figure must trace back to a primary source)
• ✗ Generate fake screenshots (hands-on screenshots must be real)
• ✗ Generate hands-on data (hands-on must be real)

5 · Conflict of interest

See about page for the full policy. Summary: members hold BTC / ETH / SOL / USDC / USDT / OKB for testing and daily use. No holdings in any project-specific token, meme, or unreleased-airdrop expectation that we recommend positively. OKX referral compensation is publicly disclosed.

6 · Corrections policy

We acknowledge we make mistakes:

1. Confirmed errors get fixed in the article body within 24 hours
2. Logged to corrections with article URL + location + old/new diff + date + reason
3. Large changes (conclusions affected) get a callout at the top of the article
4. Source corrections propagated to the source where possible

We don't delete the old version. The full trace — missed, found, fixed — is part of the transparency.

7 · No-impersonation

• We don't use OKX's brand colors, fonts, or logos
• We don't use phrases like "official download", "latest address", "tap to claim"
• Our domain doesn't carry OKX / Binance / project-similar prefixes
• We don't run lookalike social accounts

8 · Update cadence

• Pillar articles: full re-review every 3 months, immediate update on major events
• Comparative data (prices / TVL / liquidity): monthly re-check
• Corrections: realtime as discovered

9 · Feedback channels

Disagree with content / want to provide updated data / want to add a real case? File via the corrections page or email privacy@okexcex.com.