Your analytics calls it direct. It was ChatGPT.

You cannot fund a channel you cannot see.

Every quarter someone asks whether AI referrals are worth the investment, and the honest answer from a standard analytics install is that nobody knows. The traffic is there. It is just filed under direct, next to the bookmarks.

Most answer engines send visitors without a referrer, or with one that analytics does not recognise as a search source. The visit lands in the direct bucket alongside bookmarks and pasted links, so the channel looks like it does not exist. AirPulse identifies the referral patterns each engine uses and pulls them back out.

Judge it on behaviour, not volume. AI referrals typically arrive later in the buying cycle, because the engine has already answered the early questions. That is why the quality signals sit next to the session count rather than on a separate screen. Low volume with strong intent is a different decision than low volume with high bounce.

Visibility tells you whether an engine names you. AI Traffic tells you whether being named sends anyone. Both matter, and they move independently. You can gain share of answer and see no traffic if the answer resolves the buyer's question without a click, which is itself useful to know.

ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews. Coverage tracks how each engine passes referral information, which changes as they ship. Where an engine gives us nothing to work with, we say so instead of guessing.

AirPulse reads from the analytics you already run rather than asking you to add another tag to every page.

Two things. Fix the pages engines already send people to, because that traffic is arriving whether or not the page is ready for it. Then look at what those pages have in common and write more of them. The list is the cheapest content brief you will get.