ChatGPT Sent You Traffic on May 7. The Question Is What For.
On May 7, 2026, ChatGPT changed how it links to brands. Names that used to sit in footnote-style citations became clickable links inside the answer itself, pointing mostly at homepages. For most users it was invisible. For anyone supporting a brand with data, a referral source that was small enough to ignore became one worth accounting for. ChatGPT was already sending traffic before May 7. It started sending materially more.
The trackers sized it quickly. Similarweb reported ChatGPT referral traffic up about 157% week over week, homepage referrals up about 355%, and the homepage share of ChatGPT referrals climbing from roughly a quarter to roughly 60%. The elevated levels held, so this reads as a lasting change, not a blip.
That is the headline, and it is an average. An average is the one figure guaranteed to describe no single brand exactly. A 157% lift across thousands of sites tells you the tide moved. It tells you nothing about what moved for you.
The average hides something specific. ChatGPT links what it can recommend, and what it can recommend depends on what you do. A service or experience brand gets named, and the name resolves to a homepage. A catalog retailer gets a specific product recommended, and that resolves to a product page deep in the site. A publisher often is the answer, summarized inline, with no reason to click at all. Same change, three different shapes, and only your own data shows you which one is yours.
So treat May 7 as a reason to read your own numbers, not a stat to repeat. A few questions get you most of the way there.
Did your traffic actually move, and exactly when? Find your own step in the data instead of assuming the industry date applies to you. Timezone boundaries mean your change may land on the 7th or the 8th, and pinning the real date is what lets you tie the lift to a cause.
Is it more visitors, or just more activity? A referral count rises either because more people arrived or because the same people generated more hits. Compare pages per session before and after the step. Flat means net-new sessions. A shift means you are looking at something other than new visitors, and you should find out what.
Where does the traffic land? The landing page is the most informative cut you have. Homepage, product, category, article, careers. The distribution tells you what ChatGPT is sending people to, and how it shifted after May 7 tells you what the change did to you specifically.
What are you being recommended for? Read the landing pages as intent. A homepage entry is discovery. A product page is shopping. A careers page is a job seeker. The same traffic total can mean very different things, and each one points to a different team and a different conversation.
Is the traffic commercially relevant? A bigger number is not automatically a better one. Visits that land on product and booking pages carry intent. Visits to the homepage or a careers page are something else. Size the channel by what people do once they arrive, not by how much it grew.
If you saw nothing, what does that tell you? A flat line is information. It may mean ChatGPT is answering questions about you without sending a click. For some businesses that is the more important finding, and it deserves its own look rather than a shrug.
One practical note. The referrer arrives stripped to the origin, chatgpt.com, with no path, so you cannot see the prompt behind a visit. You can see everything after the click: where people land, how they behave, whether they convert. That is enough to manage this like any other channel. Set up a segment for AI referrers and watch it the way you watch organic or paid.
The trackers gave everyone the same number. Your data is telling you something more specific. The work, as always, is to go read it.
jason thompson is the CEO of 33 Sticks, an analytics botique that helps companies understand what their data is actually telling them, not what the industry average says it should. If you want a clear read on what AI referral traffic is doing to your brand, let's talk.
