Topic research tool: Find the topic. Read the competition. Keep the evidence.

Up to eight, taken from the top ten organic results. Your own domain comes out first, and so do the fourteen hosts that are never the article you are competing with: YouTube, Reddit, Wikipedia, Quora and the rest. The run then reports how many of those eight came back, so a partial read shows as a partial read.

A page that returns a block, or comes back with almost no text, gets one more attempt through a rendering service. A page that times out or fails DNS does not, because rendering cannot fix a site that is down. Anything still blocked is counted in the total and kept out of the fourteen-day page cache, so the next run tries it again.

A third-party search-data provider, in one batched request per run covering the primary keyword and every coverage term the research found. Difficulty is 0 to 100, volume is monthly searches in the country you picked, and twelve months of monthly volume come back with it. If that lookup fails the research still completes and those columns are simply absent.

No. The candidate terms and the candidate facts are both extracted from the scraped pages in code: a term has to appear on at least 30 percent of them and never fewer than two, and a fact has to be a sentence carrying a number or a citation. The model that tidies the list may only select, merge and reword what is already in it, and any fact whose numbers do not trace back to a candidate is dropped. If the model is unavailable the deterministic list ships as it is.

It deletes that brief for good, including the coverage terms and evidence already researched for it. The confirmation says exactly that, and there is no undo. You can select several and clear them at once; briefs that have already produced an article are left alone.

No, and that is deliberate. Discovery is never gated. The brief carries a compliance pre-check naming the governing bodies that apply to your site, the disclaimers that will be enforced, and any banned phrasing sitting in the brief’s own text. The publish gate itself fires on the finished draft, not on a keyword.