How the questions are written
Generate reads the brief: the brand, what it sells, the competitors you named and the market. It writes questions across seven types, so the collection does not ask the same thing seven ways.
| Type | Shape |
|---|---|
| Direct recommendation | best [category] for [need] |
| Comparison | [tool A] vs [tool B], or alternatives to [tool] |
| Use case | how do I [task], what do people use to [task] |
| Feature filtered | [category] with [attribute] |
| Problem first | the pain point, expecting a recommendation |
| Integration and market | works with [tool], or for the Indian market |
| Trust and price | is [brand] legit, [brand] pricing |
Each is written in one of three registers: formal English like a Google search, casual mobile-typed English with its abbreviations, and a spoken transcript with its filler words. Engines answer the three differently, and buyers use all three.
Your own prompts
Type a prompt and add it, or import rows from the Search Console library. A prompt can be pinned to a state or district, which changes which engines answer it and from where; see districts and languages. A prompt in Hindi or Hinglish is asked as written.
Ambiguous prompts get ambiguous answers. "Alternative for profound" draws synonyms; "alternatives to Profound for AI visibility in India" draws tools. Name the category.
Limits
A collection holds up to fifty prompts. Free has one collection of ten, Starter three of fifty, Growth ten, Agency fifty across its brands. Generate is counted per press: one on Free, twenty on Starter, sixty on Growth.
Libraries
A Search Console import lands in a library, not a collection, because a site’s top queries are mostly its own name and would measure nothing. Move the rows worth asking into a collection; the library remembers which ones were taken.