Bing SERP scraper now searches like a user (search box, fresh page, unquoted task phrases) and retries habit variants instead of treating polluted CN results as long_gap.

Date: 2026-09-01 10:20 Summary: Bing SERP scraper now searches like a user (search box, fresh page, unquoted task phrases) and retries habit variants instead of treating polluted CN results as long_gap.

[question] bing 的 serp查询，根据用户习惯修改搜索策略，尽量避免被污染

[try to solve] P0 (2026-08-31) showed CN Bing international SERPs were hijacked when queries were unlike real tool searches: quoted strings, brand moved to the end, short verbs (create/make), og slang, digit-led sizes, and one reused browser page that leaked unrelated Chinese SERPs (rural bank / 135 editor / GTA mods). Quotes did not stop hijacks.

Defaults in ops/seo/bing_serp v0.2:

First request already uses the international index (ensearch=1); no domestic-first hit.
New Playwright page per query (session isolation).
Open Bing home and type in the search box (form=QBLH); URL fallback if the box is missing.
Strip quotes; expand og → open graph; hyphenate apple touch icon; move 1200x630 off the lead; rotate create zip file → zip files online.
If organic titles miss the task tokens, or one junk host owns the page, retry up to 3 habit variants. Still dirty → draft competition_tier=unusable (strategy §3.3 D/I), never ingest as long_gap.

Did not re-scrape P0 in this turn. Next Bing batches should drop quoted/reversed-brand seed lists.

[actions]

Added ops/seo/bing_serp/query_strategy.py; wired run_bing_serp / analyze / browser_util / bing_parse / io_util
Documented §3.3 I in docs/seo/2026-08-20-long-tail-gap-strategy.md; synced Skill, funnel rule, ops README

Bing SERP scraper now searches like a user (search box, fresh page, unquoted task phrases) and retries habit variants instead of treating polluted CN results as long_gap. | Dev Logs