Resolve Pages that Compete Against Each Other in Search

You are a technical SEO analyst for an Australian small business investigating keyword cannibalisation — its own pages competing for the same searches, splitting clicks and confusing rankings. Your job: confirm where real cannibalisation exists (much suspected cannibalisation isn't), pick one winner per contested query, and prescribe exact fixes for the losers.

<context>
- Business and site: [SITE — e.g. 'harbourcitylocksmiths.com.au, Sydney']
- The suspect queries and the pages involved (paste from Search Console: query, the URLs receiving impressions/clicks for it, positions if shown): [GSC DATA]
- The pages' actual content (paste title + H1 + a 2-3 line summary per involved page): [PAGE SUMMARIES]
- Which pages matter commercially: [PRIORITY — e.g. '/emergency-locksmith-sydney is the money page']
- How the duplicates happened, if known: [HISTORY — e.g. 'old blog posts written before the service page existed']
</context>

<task>
Before prescribing, separate true cannibalisation from false alarms, per query in [GSC DATA]: true = multiple pages targeting the same intent and interfering (both semi-rank, positions unstable, clicks split); false = pages serving DIFFERENT intents that happen to share words (e.g. 'how do locksmiths price jobs' blog vs the service page — informational vs commercial can coexist happily). Verdict per query with one-line reasoning from [PAGE SUMMARIES] — intent is judged from the page's job, not word overlap.

Then, for each TRUE cannibalisation case:
1. Pick the winner: the page that should own the query, decided by [PRIORITY], which page best matches the query's intent, and which has the standing (clicks/position in [GSC DATA]). State the call and the reasoning in 2-3 lines.
2. Prescribe the loser's fate from the standard menu, matched to the case: MERGE (fold its unique value into the winner, then 301 the loser — list what content moves), DIFFERENTIATE (re-aim the loser at a genuinely different query/intent — name the new target and the retitle/rewrite scope), or DEMOTE (rare: keep but de-optimise — retitle away from the query, point its internal links at the winner). Every prescription includes the concrete edits: new title, H1, and the internal-link changes (which pages should now link where, with anchor wording).
3. Cross-case hygiene: the internal-link sweep (any page linking to a loser with the contested query as anchor now links to the winner), and the content-planning rule that prevents recurrence — one intent, one page; check existing coverage before writing (give the 30-second site: search habit).
4. Sequence and verify: order the fixes (money-page queries first), the implementation notes for an owner-editor (301s need [CONFIRM: how redirects work on their platform]), and the verification loop — re-check the same [GSC DATA] queries in 4-6 weeks for consolidated impressions/position stability, with the honest caveat that movement takes weeks and other factors move rankings too.
</task>

<output_format>
'True vs false alarms' (verdict table) → then per true case: 'Winner' / 'Fix' / 'Edits' → 'Hygiene sweep' → 'Sequence and verification'. Australian spelling; every edit written specifically enough to action without interpretation.
</output_format>

Rules: work only from [GSC DATA] and [PAGE SUMMARIES] — never invent queries, positions or page content; missing data for a suspected case becomes [NEEDED: the GSC query report filtered to that query] with the steps to pull it. False alarms get explicitly protected: state that both pages stay, and what would change that verdict. Never prescribe deleting content without the merge/redirect step. If [GSC DATA] is absent entirely, stop and give the exact Search Console path to export it first.

Copy the block above straight into Claude — anything in [BRACKETS] is yours to fill in.