You don't answer one question. You answer eight.
Behind every commercial query there are 8 to 12 sub-questions. They're the searches ChatGPT, Claude and Gemini run on Google to answer it. Atlas looks at which ones your page covers and which it leaves out.
A query is no longer one search.
When someone asks ChatGPT, Claude or Gemini a commercial question, the model doesn't search once. It breaks that query into 8 to 12 sub-questions and searches them all on Google at once. Then it composes the answer from what it finds. That breakdown is the fan-out.
Your page can rank for the original query and still be left out of the answer. Because the model doesn't read your headline: it reads the answers to the sub-questions. If your page doesn't cover them, you're not in the answer.
You rank for the question. You get cited for the answers.
Other tools give you the fan-out. Atlas tells you if you cover it.
Fan-out tools hand you a list of sub-questions, and many of them make them up with an LLM instead of observing them. Atlas does the two things that matter. It harvests the fan-out the models actually search, and matches it against your page, sub-question by sub-question. It doesn't leave you with generic homework. It tells you what you answer, what you partly answer and what you miss, with the content that closes each gap.
| Other fan-out tools | Atlas |
|---|---|
| They give you a list of sub-questions. | Matches it against your page: answer, partial or miss. |
| Many make them up with an LLM. | Captures the ones the models actually search for. |
| No provenance. | Each sub-question shows its model and its intent. |
| Generic homework. | The exact content that closes each gap. |
Atlas in three steps.
No long form. One URL and one query. We do the rest.
Structure, headings, topic coverage.
We run the URL through the same extractor AI crawlers use.
The 8 to 12 real sub-questions behind your query.
We hand your query to ChatGPT, Claude and Gemini, let them search Google, and capture the searches they ran. We add Google's People Also Ask and autocomplete. Each sub-question carries its provenance.
For each sub-question: answers, partially answers, or misses.
A visual map, and for each gap, the content that closes it. No arbitrary scores.
A map of what your page covers and what it doesn't.
The grid shows you, sub-question by sub-question, what your page answers and what it doesn't, and which sources confirm it. For each gap, the content that closes it. Plus a short diagnosis that reads the pattern. Five prioritized recommendations, not 150 generic ones.
- ●Answerscomparison of options
- ○Missesoperational differences
- ◐Partialfor small teams
- ●Answerswhat tools exist
- ●Answerspricing and plans
- ◐Partialintegrations
- ○Misseslearning curve
- ○Misseshow to migrate
Of 8 sub-questions, it answers 3, partially answers 2 and misses 3.
One tool from the Magnify Tools stack. Each looks at one thing.
Magnify Tools is Magnify's hub of free tools. Each tool answers a distinct question about your visibility.
| Tool | Question | Looks at |
|---|---|---|
| Mirror↗ | Is your brand visible when LLMs talk about your category? | Brand |
| Accent↗ | Does your international localization have technical errors? | Technical |
| Compass↗ | What do I prioritize first in my SEO roadmap? | Prioritization |
| Orbit↗ | Who actually makes up my professional network? | Network |
| Sonar↗ | What's the real intent behind your keyword? | Intent |
| Atlas | Does your page answer your query's fan-out? | Page × query |
Sonar tells you what to cover. Atlas tells you if you cover it.