Skyscraper mode turns article generation into a pipeline with a quality gate. Instead of one draft, it researches the live search results for your keyword, runs a gap analysis of what the ranking pages cover, streams a draft, then scores that draft against the same content score → the rest of the product uses — and revises until the score clears the gate, up to 4 revisions. The best-scoring draft is what you get, and the score it reports is the real engine's number, not the writer grading its own homework.
When to use it
Use Skyscraper when the article has to compete — money pages, pillar posts, anything targeting a keyword where page one is established. Use the standard mode for volume work where a good single pass is enough: Skyscraper spends more time and credits per article by design.
To raise an existing page rather than write a new one, use Improve with Skyscraper → — the same pipeline, seeded with your current page.
Run a score-gated draft
To run a score-gated draft:
- Open
Generateand set up your article as usual. - Switch the mode toggle to
Skyscraper (score-gated). - Start the run. The step tracker shows each stage: research, gap analysis, draft, then scoring and revision passes.
Read the done summary
The done summary reports the final score and how many revisions it took. A run that reaches the revision cap keeps the best draft it produced and tells you its real score rather than pretending it passed. From there, you can publish as-is, edit, or rerun with a tightened brief.
Limits and usage
- Research costs one live search-results call per run; the drafting and revision passes each consume generation credits.
- The whole pipeline runs inside a hard 285-second deadline. On very long articles, that budget favors fewer, better revisions over many shallow ones.
- Maximum 4 revision passes per run.
If something isn't working
The run ends below the gate. The run summary reports the real score. The usual causes are a very competitive keyword with a thin brief, or a section contract the draft keeps missing. Add specifics to the brief — audience, angle, what you know that competitors don't — and rerun.
The run stops near the deadline. The pipeline returns the best draft it had at the cutoff, with its score. Long-form plus maximum revisions can push against the 285-second wall; shorter target lengths finish with room to spare.
The score seems different from the dashboard's score for the same text. It's the same engine. If the numbers differ, the page changed between scoring runs — rerun the dashboard score on the current text.