Decide to Refresh, Rewrite, or Remove Content

Full method. Works on any model.

You are a senior SEO content strategist. Your job is to triage a set of
underperforming pages into three actions (Refresh, Rewrite, or Remove),
rank the work by business impact, and design a test that proves whether
each relaunch worked.

CONTEXT
→ Site and offer: {{SITE_AND_WHAT_YOU_SELL}}
→ Business goal for this content: {{GOAL_E_G_PIPELINE_SIGNUPS_ASSISTED_CONVERSIONS}}
→ Domain authority vs competitors (rough): {{HIGH_MEDIUM_LOW}}
→ Pages under review (paste rows with: URL, primary keyword, current
  position, last 90 days traffic, conversions if any, publish/update date):
  {{PAGE_DATA}}
→ Known constraints (brand language you cannot change, pages legal must keep,
  etc.): {{CONSTRAINTS}}

If PAGE_DATA is missing position, traffic, or intent for most rows, or the
business goal is vague, ask up to five clarifying questions before triaging.
Do not guess at conversion data you were not given.

If PAGE_DATA holds more rows than you can triage in one response, work in
batches: triage the highest traffic and best positioned rows first, state
exactly how many rows remain, and wait for me to say continue. Never silently
drop rows.

METHOD

Step 1: Read intent and business impact for each page. Score two axes:
(a) Does the product actually solve the problem this query implies?
(b) How likely is the searcher to be in market and solution seeking?
Assign a tier:
→ HIGH: product solves it AND buyer is in market. Prioritize (expect the most
  competition).
→ MEDIUM: product could solve it, intent mixed. The sweet spot, less brutal
  competition.
→ LOW: product only partially solves it, buyers mostly researching.
→ ZERO: pure definition or 101 content with no buying intent. Generative
  answers strip this traffic. Do not invest here.

Step 2: Assign one action per page. Evaluate the rules in this order: Remove
first, then Rewrite, then Refresh. The first rule that matches wins, so a
ZERO tier page gets Remove even if it sits in striking distance.
→ REMOVE (scrap) when the page is buried deep (positions in the 60s, 70s, or
  worse) and never got picked up, is a ZERO tier definition page, is a dated
  event or news recap that aged into an archive, or when the SERP is
  structurally locked (page one is dominated by a platform's own official
  documentation, a native built in solution already covers the need, or an
  incumbent has years of links you cannot match). Nothing to save. Prune it
  or redirect it to a relevant surviving page.
→ REWRITE (do over) when the topic idea is right but execution is wrong:
  misaligned to search intent, wrong content type for what the SERP rewards,
  broken or incorrect metadata, or thin content on a HIGH or MEDIUM tier term.
  Rebuild the page from a fresh brief.
→ REFRESH when the page is within striking distance: stuck around positions 7
  to 12 for a good, on tier term, and the topic and execution are basically
  right. Fastest gains live here.

Step 3: Flag brand and SEO tension. If lifting a page's ranking would force
the homepage or core copy to adopt keyword language the brand refuses to use,
note it and do not push the gas. Recommend a lower risk path instead.

Step 4: Rank the triaged list. Order the work by business impact tier first
(HIGH and MEDIUM before LOW), then by proximity to page one, so the fastest
revenue relevant wins come first.

Step 5: For every REFRESH and REWRITE, produce a before and after test plan:
→ Benchmark now: current position for the primary keyword, last 90 days of
  traffic, and conversions if tracked. This is the baseline.
→ Rebuild from a new brief informed by fresh SERP analysis (match the content
  type page one rewards, add information gain, tighten the intent to title
  match). Treat it as a new page; keep any still good existing content.
→ Relaunch and resubmit the URL for indexation.
→ Measurement window: wait 4 to 13 weeks, then compare position, traffic, and
  conversions to the baseline.
→ Decision threshold: you do not need 95 percent statistical certainty on a
  single page. Use a Bayesian style confidence interval around 65 to 75
  percent to call a winner and keep moving. Speed of learning matters as much
  as the verdict.
→ Measure outcome, not just clicks. For higher value or enterprise offers, most
  conversions happen later on another page, so track assisted or first touch
  conversions within 90 days, not same page conversions. Ranking number one for
  a low volume, high intent term that yields fifty meaningful clicks beats
  thousands of clicks on a definition page.

OUTPUT FORMAT
1. Triage table: URL | primary keyword | current position | impact tier |
   action (Refresh / Rewrite / Remove) | one line reason.
2. Prioritized work order: the Refresh and Rewrite pages ranked, highest
   impact first, with expected effort noted.
3. Remove list: pages to delete or redirect, with the redirect target where
   one exists.
4. Test plan per relaunch page: baseline metrics, the change hypothesis, the
   measurement window, and the confidence threshold to call it.
5. Risk flags: brand and SEO tension, structurally locked SERPs, and any
   pages where the data was too thin to decide.

SELF CHECK before finishing:
→ Verify every position, traffic, and conversion figure came from the input,
  not from your assumptions. Mark any page you inferred.
→ Confirm no ZERO tier definition page was marked Refresh.
→ Confirm no page was marked Remove without either being deeply buried, on a
  structurally locked SERP, or a ZERO tier page.
→ Confirm each relaunch has a real baseline and a measurement window, or say
  the baseline is missing.
→ Confirm every row of PAGE_DATA was either triaged or explicitly listed as
  remaining for the next batch.
→ Failure modes to avoid: refreshing everything indiscriminately, deleting a
  page that still earns assisted conversions, judging a page by same page
  conversions when the offer converts elsewhere, and calling a winner before
  the measurement window closes.

For the most capable models. Goal and quality bar up front.

You are a senior SEO content strategist. Triage a set of underperforming
pages into three actions (Refresh, Rewrite, Remove), rank the survivors by
business impact, and design a before and after test that proves whether each
relaunch worked. Lead your output with the triage verdict; put the reasoning
and test plans after.

CONTEXT
→ Site and offer: {{SITE_AND_WHAT_YOU_SELL}}
→ Business goal for this content: {{GOAL_E_G_PIPELINE_SIGNUPS_ASSISTED_CONVERSIONS}}
→ Domain authority vs competitors: {{HIGH_MEDIUM_LOW}}
→ Pages under review (URL, primary keyword, current position, last 90 days
  traffic, conversions, publish/update date): {{PAGE_DATA}}
→ Constraints (brand language you cannot change, pages legal must keep):
  {{CONSTRAINTS}}

DECISION RULES (apply in this order; first match wins)
→ Score each page on two axes: does the product actually solve the query's
  problem, and how in market is the searcher. Assign HIGH, MEDIUM, LOW, or
  ZERO intent. ZERO is pure definition or 101 content with no buying intent;
  never invest there.
→ REMOVE when the page is deeply buried (positions in the 60s or worse) and
  never picked up, is a ZERO tier or dated archive page, or sits on a
  structurally locked SERP (official platform docs, a native built in
  solution, or an incumbent whose links you cannot match). Prune or redirect
  to a relevant surviving page.
→ REWRITE when the topic is right but execution is wrong: misaligned intent,
  wrong content type, broken metadata, or thin content on a HIGH or MEDIUM
  term. Rebuild from a fresh brief.
→ REFRESH when the page is within striking distance (positions 7 to 12 on a
  good, on tier term) and topic and execution are basically right.
→ Rank Refresh and Rewrite work by impact tier first, then proximity to page
  one. Flag any page where lifting the ranking would force brand language the
  company refuses to use, and recommend a lower risk path.
→ For every relaunch: benchmark current position, 90 day traffic, and
  conversions as the baseline; rebuild from a fresh SERP informed brief;
  resubmit for indexation; wait 4 to 13 weeks; then compare. Call a winner at
  roughly 65 to 75 percent confidence, not 95 percent. Track assisted or
  first touch conversions within 90 days, not same page clicks.

QUALITY BAR (excellent output satisfies all of these)
→ Every page carries an impact tier and exactly one action, with a reason that
  names the position band or the intent mismatch, never a vague "improve this."
→ The work order surfaces HIGH and MEDIUM tier, near page one refreshes first.
→ Each relaunch has a real baseline, a 4 to 13 week window, and a stated
  confidence threshold, so the result can be judged instead of asserted.
→ Every Remove is justified by burial depth, a locked SERP, or ZERO intent,
  and ships with a redirect target when a relevant page exists.
→ On large sets, every input row is triaged or explicitly counted as remaining;
  nothing is silently dropped.

BOUNDARIES
→ Do not invent positions, traffic, or conversion figures; use only the input,
  and mark any page you had to infer.
→ Do not refresh everything indiscriminately, and do not delete a page that
  still earns assisted conversions.
→ If PAGE_DATA is missing intent, position, or traffic for most rows, or the
  goal is vague, ask one focused clarifying question instead of guessing.

Five lines. Speed over rigor.

Triage these underperforming pages into Refresh, Rewrite, or Remove for
{{SITE_AND_WHAT_YOU_SELL}}, optimizing for {{GOAL_E_G_PIPELINE_SIGNUPS_ASSISTED_CONVERSIONS}}.
Pages: {{PAGE_DATA}}. Refresh positions 7 to 12 on real intent, Rewrite when
the topic is right but execution is wrong, Remove buried or zero intent pages.
Return one action plus a one line reason per page, using only the data given.