Turn a stack of real customer support questions into a long-tail keyword cluster worth writing for

You are helping build a long-tail keyword cluster from real customer language, not from a keyword tool's suggestions — the raw material is actual questions people asked support or sales, in their own words.

RAW QUESTIONS
'Can I use this if I'm already on the annual plan?' / 'Does the API rate limit reset at midnight UTC or my local time?' / 'What happens to my data if I downgrade?'

PRODUCT OR SERVICE CONTEXT
B2B API metering tool that tracks usage-based billing for SaaS companies.

EXISTING CONTENT THAT ALREADY COVERS SOME OF THIS
A billing FAQ page, a plan comparison page, and a docs page on rate limits — none specifically address downgrade data retention.

STEP 1 — GROUP
Read through the raw questions and group them by the actual underlying question being asked, not by surface keyword overlap — two questions phrased completely differently can be the same underlying intent, and two questions sharing a keyword can be asking genuinely different things. Name each group by the real intent, not a generic label.

STEP 2 — EXTRACT LONG-TAIL PHRASING
For each group, extract the specific long-tail phrasing actual customers used — preserve their real word choices, including any non-obvious terms, misspellings that suggest a real search pattern, or informal phrasing, rather than "cleaning it up" into generic marketing language. Note where the natural phrasing differs from what a keyword tool would suggest.

STEP 3 — CHECK AGAINST EXISTING CONTENT
For each group, check whether existing content already covers it adequately, partially, or not at all. Do not assume a keyword is a gap just because you don't see it named explicitly — check whether the underlying question is actually answered anywhere in the existing content, even under different wording.

STEP 4 — PRIORITIZE
Rank the groups by a combination of how often the underlying question appeared in the raw data and how clearly it signals someone close to a decision (comparing options, asking about a limitation, asking "can I") versus early-stage curiosity.

OUTPUT FORMAT
A table with columns: Question Group | Real Customer Phrasing (2-3 examples) | Coverage Status (covered / partial / gap) | Priority (high/medium/low) | One-line reason for the priority.