Query fanout prompt: plan useful research questions

A query fanout prompt can generate possible research questions. Those suggestions are not executed searches. Use the template for planning, or Native fanout to inspect the exposed queries of a real API web-search run.

Use this for an initial topic plan before checking demand and evidence. The template belongs in a chat or your own workflow; this site's tool accepts only a short topic.

Choose the right kind of resultNative fanout observes API searches. Search Ideas supports planning without web search.Open the free tool

First decide: planning or observation?

If you want possible reader questions, a planning prompt is enough. Its output shows what the model suggests. A plausible list establishes neither a prior web search nor human search demand.

If you want to see what a provider actually searched for, you need a run with a search tool enabled and its response. Native fanout on ai-fanout.com reports only exposed query strings and sources. Missing queries stay missing; a planning prompt does not fill them in.

For the distinction from the consumer product, read the OpenAI search-query guide.

Search Ideas is this tool's separate second mode. It generates ten modelled research directions without web search. Use it as a starting point and check the suggestions against your own evidence.

Which output do you need?

Choose by the task, rather than by the length of the generated list.

Task	Suitable route	What remains to check
Collect possible reader questions	Planning prompt or Search Ideas	Evidence and demand are still unverified.
Inspect exposed searches	Native fanout with OpenAI or Gemini	The API run, visible fields and source scope.
Improve a page	Match questions to existing content	A distinct reader task and the evidence it needs.

A template for a reviewable research plan

Replace the three placeholders with the topic, reader decision and market. Ask for research questions rather than finished facts. The evidence column shows which tasks still require research.

The template below is owner-written editorial material. It is not a saved provider response or a list of hidden searches. Copy it into your own chat if useful; reading this page does not execute it.

Then review each question: does it serve the audience, does an existing page already answer it, and can you support it with current evidence? Combine wording variants and remove questions that do not support a relevant decision.

Turn that plan into a page decision with the query fanout SEO workflow.

Planning prompt to copy

Topic: [one topic]
Audience and decision: [who needs to decide what]
Market and language: [market, language]

Suggest up to ten distinct research questions that could help this reader.
Separate wording variants from genuinely different information needs.
For each question, name the evidence needed and whether an existing page
could answer it with a section. Mark every item as a modelled suggestion.
Do not claim these are queries executed by Google, ChatGPT or Gemini.
Do not invent sources, search volume, difficulty or rankings.
Return: question | reader task | required evidence | page or section | status.

Choose short inputs for the free tool

The tool accepts 2–60 characters and at most 160 UTF-8 bytes. URLs, files and multiline briefs do not belong in the input field. Enter one topic and choose country and language in their separate controls.

For a software decision, CRM for small agencies is a possible input. For an introduction, try SEO for AI search; for a comparison, Ahrefs vs Semrush. These are examples of acceptable topics, not predicted output queries.

Choose Native fanout to observe exposed API searches or Search Ideas for a modelled plan. Complete the security check and start the run yourself. Afterwards, export only the questions you want to investigate further.

The next step is covered in the CSV and JSON export guide.

Before and after: a bounded research assignment

The exercise starts with a prompt asking for every keyword and every required page. It leaves reader, market, existing URL and acceptance criteria open. A long answer would not yet produce a useful page assignment. The revised prompt concerns a two-person team's German-market SEO software decision; /seo-tools is its fictional existing comparison page.

The revision limits the task to six research questions with reader job, existing section and required evidence. It requires modelled labels, open demand fields and no invented prices or rights. The download contains both prompt versions. Its follow-up review questions are original editorial material, not recorded model output.

Our selection retains the team decision, merges pricing variants into one plan section and separates the CSV-import task. An unrelated SQL question is excluded. The stop rule applies when current plan, precise seat scope or export permission is absent: research first, then write a bounded claim. More wording variants would not close that gap.

See the resulting concrete assignment in the product-comparison brief.

Use the long version in your own planning workflow. This public tool still receives only a short topic; it does not execute the downloaded brief. If you later observe actual API queries, retain them as a separate new evidence source rather than part of this editorial exercise.

Common questions

These short answers cover recurring questions from tool use and source review. They are not claimed search-volume data.

Can a prompt show the real Google AI Mode queries?: A generated list is not a recording of a Google session. Google publicly describes query fan-out; this tool does not read private AI Mode queries.
Should I paste the whole template into the tool?: No. Enter a short topic here. The template is for separate planning in your own chat or workflow.

Sources used

These sources support the functions and limits described here. They do not prove claims beyond that scope.

Original annotated workflow, controls, prompts and export exercises
Original diagrams and editorial annotations; only the previously approved normalized OpenAI records. Synthetic selection files use actual producers and independent CSV readback, not new provider output or Excel/Sheets import. Own dated public-page checklist makes indexing and outcomes explicitly unproven.
Open source
ai-fanout.com input, selection export and local-history implementation
The input controls, separate native/modelled modes, three export scopes, query-selection CSV fields, and browser-local history are described from the site's actual implementation. The planning prompt and input examples are original editorial instructions, not provider output or measured keyword demand.
Open source
Optimizing for generative AI features on Google Search
Normal SEO foundations remain relevant; pages made primarily for fan-out variations are not the accepted content model.
Open source
Direct OpenAI and Gemini native-fanout implementation
One direct provider request, strict provider allowlist, provider-exposed query/source parsing, honest zero-query output, CAPTCHA-first shared atomic quota and budget accounting, no raw server archive and separate public flag. Bounded live production fixtures passed for OpenAI and Gemini on 2026-08-26.
Owner record