model · available

Query Fanout 1

Give it a question someone would put to an AI assistant, and it returns the web searches that assistant is likely to run before it answers. We trained it on 3,217 searches we watched ChatGPT, Claude and Gemini actually make. If you care whether your pages get cited in AI answers, those searches are what decide it.

Try it in the browser

spec

model id: query-fanout-1
openrouter id: meragpt/query-fanout-1
context: 4,096 tokens
max output: 160 tokens
input price: $0.05 / 1M tokens
output price: $0.15 / 1M tokens
repeatability: varies between calls
max request: 2,000 characters

built for

Generative-engine optimisation: finding the searches that decide whether your page is cited in an AI answer.
Content planning, where the queries a question triggers are a better brief than a keyword list.
Competitive research: seeing which rivals an assistant pulls in when someone asks about your category.

not for

Telling you exactly what ChatGPT or Gemini will search. Ask the same engine twice and it writes different queries both times, so nobody can promise you that, us included. Treat the output as the ground a question covers.
Keyword volume, difficulty or ranking data. You get the searches themselves and nothing about how valuable they are.
Questions with no buying intent. We trained it on how people shop across forty commercial categories.

calling it

One question, one list.

curl https://meragpt.com/v1/fanout \
  -H "Authorization: Bearer $MERAGPT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "query-fanout-1", "question": "..."}'

Send one question of up to 2,000 characters and get back up to 6 searches. Nothing is split: a question is one generation, and the list is deduplicated as a whole, so two near-identical searches collapse into one before you see them. Naming the brands in the category with brands is optional and sharpens the result. There is also an OpenAI-compatible /v1/chat/completions shim, where the searches arrive newline-joined in the message content and as an array under meragpt.queries.

evaluation

How it measures up

Tested on whole product categories the model had never seen during training, against the searches real assistants ran on those questions. Scoring compares what a search is looking for rather than its exact wording, because an engine never writes the same query twice.

coverage	0.563	of the searches a real engine ran, by token overlap
agreement ceiling	0.75–0.91	what one real engine run scores predicting another
queries returned	4.2	mean, capped at 6
set diversity	0.364	real engine fan-outs measure 0.394

limits

What it does not do well.

Expect the right territory rather than the exact wording. Engines rewrite their queries every time they answer, so nothing can match one run word for word and still match the next.
Check any brand name it gives you. At this size the model will sometimes invent a competitor that sounds plausible and does not exist.
We trained it on buyer questions in forty commercial categories. Support, medical, legal and local-services questions are untested.

disclosure

Query Fanout 1 is our own model, built and trained in house for this one task. The weights are proprietary and are not published. What is published is the measurement: every number on this page comes from a test run on data the model had never seen in training, not from a demo we liked. Model ids are stable and are never renamed — a change in behaviour ships as a new id, so an integration pinned to query-fanout-1 keeps the model it was tested against.

Calling it from your own code is in the API docs. Pricing and the credit model are on the homepage.