Understand how your game appears in AI answers.
Keep the questions separate: what an answer says, which sources it names and whether anyone follows through to your site.

Three different things to observe.
A mention
Was the game named, and was the description accurate? Record the exact question and date.
A citation
Which source was linked or attributed? Keep the source and the answer together.
A visit
Did your available analytics record a referral? Name the measurement window and its limits.
Read the existing explanations.
Reading the Search Console AI report
The existing explanation of the report and its limits. Check the article date against the interface you use.
Questions about appearing in ChatGPT
A dated guide to an evolving topic, not a guaranteed route into an answer.
Understanding llms.txt
The existing explainer; separate a proposed convention from demonstrated results.
Readable pages, useful explanations
The earlier article on text and presentation, with its original context.
AI visibility and indie games
The existing discussion of visibility and developer outcomes. Treat its broad conclusions as questions to examine, not sales evidence.
Make an experiment you can revisit.
Write down the question
Use a small set of relevant questions and keep their wording with the record.
Record the conditions
Save the date, service and settings available to you. Note which parts you cannot control.
Separate observation from interpretation
A changed answer is something to investigate, not proof that one edit caused it.
Keep outcomes distinct
Publisher visits do not establish developer wishlists or sales. Use actual developer data for those claims.
These pages are not all personal experiments.
Some explain documentation and research; others discuss observations from Crit Pick. The dated articles keep that context. Proposed tests with and without LLMs remain investigations to carry out, not completed results.