AI Model Wiki

    Skip to main content

    Update Policy

    Last updated: February 2026

    Every page on AI Model Wiki has a refresh schedule. This page documents those schedules and the staleness thresholds that determine when a badge turns yellow or red.

    Refresh cadence by page type

    • Model pages — verified every 30–60 days, immediately on a major version change. Pricing on top-20 models re-checked every 14 days. See Pricing Data Policy.
    • Comparison pages — refreshed when either underlying model is updated, otherwise every 60 days
    • Tool directory entries — every 60–120 days
    • Blog posts — refreshed at the editor's discretion; freshness tier shown on every card
    • News items — usually not refreshed; news is timestamped at publication and remains historical
    • Regulation pages — every 90 days, immediately on major legislative changes
    • Benchmarks — synced to the source leaderboard's update cadence; we re-pull at least monthly. See Benchmark Methodology.
    • Calculators — pricing inputs follow the Pricing Data Policy; methodology revisions happen quarterly
    • Methodology & policy pages (this one) — reviewed every 6 months or on operational change

    Staleness tiers & visible badges

    Every model page shows a colour-coded last verified badge:

    • Verified Xd ago — within the fresh window for the content type
    • Last verified Xd ago — past fresh threshold, still within stale threshold
    • Needs update (Xd) — past the stale threshold; treat numbers with caution and confirm against the linked source
    • Needs verification — never editorially confirmed

    Internal freshness audit

    We run a freshness audit script weekly. It groups every content table by age tier and produces a report flagging anything past its stale threshold. The editorial team works through the report each week.

    Major version changes

    When a developer ships a major new version (e.g., GPT-5 → GPT-5.5, Claude Opus 4 → Opus 4.5), we treat it as a trigger for an immediate refresh of:

    • The model's own page
    • Every comparison page that includes the model
    • Every "best AI for" page where the model is recommended
    • The benchmark hub if new benchmark numbers are available

    What triggers a re-verification

    • The scheduled refresh interval elapses
    • A reader reports outdated info via the in-page link
    • A developer publishes a relevant announcement
    • A benchmark we cite updates
    • A pricing change is announced

    Reporting outdated content

    See Corrections Policy. Reports route to the editorial team and feed into the verification queue.