A content backlog built from what the niche asks

Most content calendars are built from what the team wants to say. This loop builds one from what the niche is already asking, and keeps the link to every place it was asked, so the backlog can be argued with.

Use it when you are planning content for a niche and want the list to come from evidence rather than a brainstorm.

You need a niche definition, a context file for your product and audience, and the sources to mine.

You get a scored backlog. Each idea has the question in the audience’s own words, the links where it was asked, a count of distinct places it appeared, and scores for gap and fit with a sentence each.

The gate is nothing, because it publishes nothing. It hands you a list.

The three scores

• Evidence. How many distinct sources asked this, counted, not estimated. Two threads in the same community by the same person count once.
• Gap. Whether the current top answers are thin, dated, or missing the part people keep asking about. Read the answers; do not guess.
• Fit. Whether answering it honestly leads anywhere near what you sell. High evidence with zero fit is a traffic idea, not a pipeline idea. Label it that way.

Goal prompt

Read <niche-file> and <context-file>. Mine each source in <sources-file>:
community threads, video titles and comments, search autocomplete for each
seed term, and the questions answer engines suggest as follow-ups.

Collect every question or request in the audience's own words, with its URL
and date. Cluster questions that ask the same thing. For each cluster, write
one idea: a working title, the best verbatim phrasing, every URL, and a count
of distinct sources.

Score each idea: evidence (the count), gap (read the current top answers and
say what they miss), and fit (does answering it lead toward <product>). Give
one sentence of reasoning per score. Rank by evidence, then gap, then fit.

Write to <output-file> with the collection date. Never add an idea with no
source URL. Never invent a search volume or a view count.

Verifier prompt

Open the output file in a fresh session and load the URLs.

FAIL if any idea has no evidence link.
FAIL if any evidence count does not match the number of distinct sources listed.
FAIL if any verbatim phrasing does not appear at one of its linked URLs.
FAIL if any gap score has no reasoning naming what the current answers miss.
FAIL if any search volume, view count, or traffic figure appears without a
named tool and date.

Otherwise PASS. On FAIL, print the idea and the reason.

Stopping condition

Every source was mined or has a recorded error, every idea in the backlog has at least one evidence link, its count matches its links, and the verifier printed PASS.

What it may and may not do

May: read public communities, video pages, autocomplete, and answer engine suggestions; write the backlog.

May not: log in to private groups, invent volume data, add ideas from its own knowledge without a source, or write the content itself.

What this will not tell you

Loud questions are not always valuable ones. A niche where everyone asks the same beginner question may already be answered well, and the gap score exists for that reason.

Freeze the backlog once it is scored. Re-run it on a schedule and compare the new list to the old one, instead of editing the frozen list as you go.