Evidence, not claims
AI fanout examples you can inspect yourself.
These examples show which web searches the OpenAI API exposed in four specific runs. Each case keeps the input, model, locale, query strings, sources, token use and limits together.
- 4
- OpenAI runs
- 16
- exposed queries
- 1
- fixed protocol
Observed
How to read the examples
A single run is a dated observation. The pages separate three layers so that the conclusion never grows beyond the evidence.
- 01
Returned by the provider
Exact query strings, search-action sources, model, locale, status and usage.
- 02
Interpreted by us
A comparison of the visible data: what stayed stable, what changed and which reader job the branch serves.
- 03
Not established
No claim about hidden ChatGPT searches, private reasoning, overall model quality or future runs.
Choose the question you want to investigate
The four cases serve distinct jobs. They are not interchangeable SEO pages.
- 01How to compare OpenAI and Gemini fanout locally
How can you compare OpenAI and Gemini for “best SEO tools” without inventing missing evidence?
- 02How much does country change fanout queries?
Does the country setting change the queries for the same keyword?
- 03Why does the same keyword produce different queries?
Can the same setup return different fanout queries?
- 04Which sources appear for AI comparison questions?
What does an AI web-search run look for when comparing Ahrefs and Semrush?
JSON · CSV · SCHEMA
Download the data and method
The normalized OpenAI observations are available as JSON and CSV. The schema records the fields and versions. Sources belong to the full search action; the provider does not necessarily map them to one query inside that action.
Next: repeat the same topics over time
A preregistered set of twelve informational, comparison, commercial, troubleshooting, implementation and local questions is planned for monthly repetition. Only completed, reviewed runs become published results. The plan itself is not a result.
Open the observation planTest your own topic
Run a free observation with OpenAI or Gemini. Your result starts in this browser, where you can save, compare or export it.