Freshness window web-run/freshness-window

What it is

The third parameter of a web.run line is a number in days: slow|Nike Pegasus 41 günstig kaufen Deutschland Preis 2026|7. It limits how old a result may be. Measured 22 Aug 2026 per category: news (K5) 1 day, local opening hours (K1-02) 1 day, prices (K3) and news background 7 days, local and short searches (K1-01, K13, K2-01, K11-02) 30 days, product recommendations with a model list (K2-02, K2-03) 90 days, and — the most striking case — stable facts 3650 days (ten years): a person's biography (K6-01: fast|Sam Altman biography Loopt Y Combinator|3650) and a device's specifications (K7-02: fast|iPhone 17 technical specifications Apple|3650). Evergreen queries without a search carry no window.

External evidence

• Suganthan Mohanadasan, 21 Aug 2026: windows by topic (stocks 2, sport 7, commerce 30, financial advice 90, Reddit 3650 days). Divergence: in our data 3650 is not for Reddit but for stable facts (biography, device specs). The window follows fact stability, not the platform.

For visibility

The window is why a stale price comparison does not appear in ChatGPT even if it ranks in Google. A product page whose last visible change is older than 30 days does not compete for price and short searches at all. Conversely, for stable facts ChatGPT opens a ten-year window — there, permanence counts, not freshness. What ChatGPT uses to determine age (publication date in markup, pub_date in the index, modification date) is not established; entries[].pub_date in the results suggests the index keeps a date per hit.

External sources

mohanadasan-2026-08-21

Change history

No event has touched this object yet.

Open

• Is the window a hard filter or a weighting? Its form as a pipe parameter suggests a filter; it is not proven.
• What does the model base the window on? Our data points to the stability of the fact sought (news 1 day, biography 3650 days), not the platform or a signal word.

Related

Query fan-out web.run syntax (pipe format) Domain restriction