ChatGPT Traffic Jumped 60% Overnight. Here's What It's Really Buying.
ChatGPT referral traffic jumped 60% overnight. The clicks are not the win. They are the training data OpenAI needs to price the ads it is about to sell you.
ChatGPT’s traffic spike is real. It is also the least interesting thing that happened in May.
On May 7, ChatGPT started hyperlinking brand names straight to company homepages. Plain bold text became a clickable link. Referral traffic to monitored brands jumped roughly 60% overnight and held.
Most people read that as a traffic story. Based on the calendar, I’m reading it as an ad-tech story.
What actually happened
Start with the size of it, because the size is what makes this real instead of anecdotal.
In Profound’s monitored set, daily OpenAI referrals roughly doubled starting May 7 and stayed there. Homepage share of those referrals went from around 3.5% to around 24% in a week. Brand front doors went from a rounding error to one in four referral clicks.
This is not one dataset talking to itself. Similarweb’s clickstream panel measured total ChatGPT referrals up about 158% week over week, with homepage traffic up 355%. Qwairy studied more than 140,000 responses and found inline brand links jumped from 0.4% of answers to 6.2% in a single day. Three independent vantage points. One event.
Josh Blyskal and the Profound team caught it in the wild first (props). One of the most important observations in AI search this year.
Real, attributable brand clicks where there used to be almost none. If that were the whole story, it would be a good one.
It is not the whole story.
Now look at the calendar
May 5: OpenAI opened its self-serve Ads Manager. CPC bidding. A conversions API. No minimum spend.
May 7: brand hyperlinks rolled out.
Two days. You do not need a leaked roadmap to read that sequence. A relevance-weighted CPC auction needs one thing above all else: a clean record of what people click. The branded link update is how OpenAI started collecting it, at scale, across every brand and every query at once.
Every clickable mention now throws off a signal. Which brand got clicked. On which question. In which context. At what rate against the other brands in the same answer.
That is the raw material of an ad ranking system.
Why the clicks matter more than the traffic
Google’s ad machine was never built on impressions. It was built on learning which results earn clicks.
CTR is not a reporting metric there. It is the input that decides what gets shown, where, and at what price. Impressions were the warm up. Clicks were the unlock.
OpenAI is running the same play in compressed time, and it is not hiding it. The Ads Manager that went live on May 5 runs a relevance-weighted second-price auction. That is the published mechanism. It means a tighter, more relevant ad on a smaller bid beats a lazy ad with a bigger budget. Relevance gets weighted. To weight relevance, you have to know what people actually click. The branded link update is the meter that reads it.
So the two events are one event. The traffic win is real. It is also the byproduct. The product is the click data itself.
Who won, who sat out, and why
The category split makes sense the moment you stop counting traffic and start reading intent. In Profound’s data, B2B SaaS referrals jumped more than 200%. Fintech rose around 60%. E-commerce stayed roughly flat.
ChatGPT is minting clicks in the categories where people ask for a company by name. Best tool for. Alternatives to. Who should I use for. Purchases with a transactional path route through the shopping surface instead, so retail sits the link auction out for now.
The brand auction is getting built first. If yours is a category where buyers ask ChatGPT for a name, you are already in the training set whether you planned to be or not.
Three moves while the window is open
The free-traffic window is open right now. Treat it like inventory you can earn before you have to rent it.
Make ChatGPT a named channel in your reporting. Add chatgpt.com and openai.com as referral sources today. Google made this easier on May 13 by adding a native AI Assistant channel to GA4 that classifies ChatGPT, Gemini, and Claude traffic with no custom filters. One caveat. A large share of AI visits arrive from mobile apps that strip the referrer and land in your direct bucket, so the reported number undercounts the real one. Tag what you can. Assume there is more.
Treat your homepage like a cold landing page. One in four AI clicks now lands on your front door from someone who knows nothing but your name. They arrive with the one sentence ChatGPT wrote about you and about five seconds of patience. If they cannot tell what you do and who it is for in that window, you earned the click and handed it right back.
Win share of answer while the clicks are still earned. This is the Answer Ownership move. Every click you earn organically today is a data point in the auction that prices your ads tomorrow. The brands that own the answer now set the floor everyone else has to outbid later. The ones that wait will pay to buy back visibility they could have had for free.
The part that should make you uncomfortable
OpenAI pulled in over $100 million from a six-week pilot before any of this went self-serve. The stated target is $2.5 billion this year and $100 billion by 2030.
Those numbers do not come from showing ads. They come from pricing them well. And pricing them well requires exactly the click data the May 7 update just switched on.
You are not the customer of that system yet. Right now you are the free traffic. Every click you earn today becomes the case study OpenAI uses to price your ads tomorrow.
The window where AI mentions are earned, attributable, and free does not stay open. Own the answer while it is still organic.
Great work to Josh Blyskal and the Profound team. This is the most important shift in the category today.