About LaunchAutopilot

Launch sites measure attention. They can tell you how many people upvoted you on a Tuesday. None of them can tell you whether the thing that actually decides most software purchases now, an assistant answering a buying question, has started naming you.

How it works

Before you launch we freeze a set of questions your buyers genuinely type, and ask ChatGPT, Perplexity and Gemini each of them many times over. We record how often your product comes up and which sources the answers cite. Your name never appears in the question.

You launch in a curated weekly cohort. Thirty days later we ask the same frozen questions again, alongside control questions in your category where you would not be expected to appear. If the controls rose too, the whole category rose, and we subtract that from your result. Then we publish what changed, whichever direction it went.

What we will not do

These are rules in the code, with tests, rather than promises on a page.

• No AI “rank”. Engines do not rank products. You come up in some answers and not others, so we show a mention rate with its range, and say “not enough data” when that range is too wide to mean anything.
• Paid never outranks earned. A Launch Pass picks which week you launch in. A featured spot is a fixed, labelled position above the board that takes no part in the ranking. Both render as sponsored links. Neither touches editorial review.
• No badge-for-link trades. Badges are earned and optional, and your listing never depends on whether you embed one.
• No hidden rejections. Every rejection and its reason is published.
• Measurement is frozen before any intervention and always includes control questions, so a result cannot be produced by changing the question.

Reddit, specifically

Part of the system drafts Reddit contributions for makers. It uses the maker’s own account and nobody else’s, every single item is approved by that maker before it is posted, and promotional posts carry a disclosure. There are no account pools, no karma farming, no vote manipulation and no attempt to evade detection. A plan that needs any of those is not a plan we will run.

Who runs it

LaunchAutopilot is built and operated by OceanLab Technology. The code is public at github.com/xocketlabs/launchautopilot, including the scoring, the guardrails and their tests, so the rules above can be read rather than taken on trust.

The first cohort has not opened yet and nothing has been measured so far. Any figure you see on this site today is labelled as an example, because that is what it is.