Query Fan-Outs or Prompt Volume: Which Should Drive Your Content Plan?
Prompt volume tells you which questions matter. Query fan-outs tell you what the AI engine actually searches for while answering. How to combine the two so an agent writes the right page for the right query.
Teams that move past keyword research for AI search usually pick one new signal and build everything on it. Some sort their plan by prompt volume. Others chase query fan-outs because they show what ChatGPT actually searched. Both camps end up with gaps, because the two signals answer different questions.
Prompt volume decides which conversations are worth winning. Fan-outs decide what to write so you get retrieved in them. An agentic content workflow needs both, in that order.
What each signal measures
Prompt volume estimates how often a question gets asked. In Promptwatch, it's built from the keywords attached to a prompt, with monthly AI search volume weighted five times, Bing 1.5 times, and Google 0.5 times, shown as a band of one to five bars. Difficulty sits next to it, using the last 30 days of keyword competition and the authority of sources AI already cites.
Query fan-outs are the sub-queries an AI assistant runs on the web while composing an answer. The user asks one question; the engine searches for something shorter and more specific, sometimes several things. Promptwatch shows fan-outs per model, so you can compare how ChatGPT and Gemini break down the same prompt, and see the terms and themes they reveal.
One signal works at the level of the conversation. The other works at the level of retrieval.
Why volume alone writes the wrong page
Say a prompt has strong volume: "what's the best project management tool for a 20-person design agency?" A volume-led plan writes a page that answers that sentence.
But the engine doesn't search for that sentence. It might search "project management software design agencies" and "asana vs monday for agencies." If your page is built around the long conversational question, it can miss the shorter query that actually does the retrieving. Promptwatch's own fan-out research makes this concrete: ChatGPT's average fan-out query length dropped from about 117 characters in early December to around 53 by April. The engine searches more like a person typing keywords.
The fix is to keep volume as the reason a page exists, and use fan-outs to decide its headings, title, and scope.
Why fan-outs alone chase noise
Fan-outs without volume have the opposite problem. Every tracked prompt produces sub-queries, including prompts nobody asks. If an agent writes a page for every fan-out it sees, you get a long tail of pages answering searches that only happened because you ran a test prompt.
Fan-out counts also move. The same Promptwatch research found ChatGPT averaged about 2.15 distinct searches per answer in early December, around 1.84 by early March, and exactly 1.0 when the data resumed in April. A plan tied to fan-out volume shifts every time the engine changes behavior. Volume and difficulty are steadier inputs for deciding what to work on.
The combined workflow
- Pick prompts by volume and difficulty. Volume above one bar, difficulty you can realistically beat, intent tagged commercial or transactional where possible.
- Check the content gap. Promptwatch's Content Gap view shows which of those prompts have no matching page on your site.
- Open the fan-outs for each gap prompt. Note the sub-queries per model. Where ChatGPT and Gemini search differently, decide whether one page can cover both or whether you need two.
- Brief from the fan-outs. Use the fan-out phrasing for the title and H2s. Front-load the entity and category, as the research recommends, since short queries match the start of a heading best.
- Prefer focused pages. When a prompt splits into distinct sub-queries ("X vs Y," pricing, alternatives), a cluster of focused pages usually beats one long page that half-covers each.
- Publish and recheck. After the page is live, check whether it's cited for the parent prompt and whether new fan-outs appeared.
Steps 3 to 5 are where an agent saves the most time. Promptwatch's Content Agent writes against visibility gaps from the Knowledge Base and publishes to WordPress, Webflow, or Framer, and fan-outs can be exported to CSV if you'd rather brief a human writer.
A free way to look first
If you want to see fan-outs before setting up tracking, Promptwatch offers a free GEO Inspector Chrome extension that shows what ChatGPT searched, which sources it fetched versus cited, and the reasoning trail. It runs locally and works without signing in. It's a good way to sanity-check a handful of prompts before building a plan.
The answer
Volume drives the plan. Fan-outs drive the page. A tool that shows only one of them leaves you guessing about the other. Promptwatch shows volume, difficulty, and per-model fan-outs on the same prompt, which is why it ranks first on our agentic SEO rankings for content planning. Essential is $95/mo for 50 prompts; the scheduled Content Agent that turns the plan into published pages is on Business at $579/mo.