E-commerce, SaaS, fintech, auto, health, beauty, astro — if any part of your traffic depends on ChatGPT, this is the pattern you need to understand, not just the headline.
If you track referral traffic from ChatGPT, you’ve probably seen a strange story play out over the last year: a sharp decline, some tense months, and then — a rebound that most “AI is killing your traffic” posts never mention.
Both halves happened. Neither half is the full story. And the sequence itself is the actual lesson.
Act One: Why Clicks Started Disappearing (2025)
Through 2025, ChatGPT increasingly behaved like it was trying to answer questions completely inside the conversation, rather than sending people elsewhere to find the answer themselves.
Alongside that shift, the platform also narrowed which sites it was willing to name. Independent analysis of over a billion ChatGPT citations found that in July-August 2025, referral traffic to a broad sample of branded websites fell by roughly half, while citations to Reddit and Wikipedia surged — Reddit up 87%, Wikipedia up 62% — with the top three cited domains alone controlling close to a quarter of all citations.
Wikipedia and Reddit still lead ChatGPT’s citations today — this part of the pattern never fully reversed.
Two more forces compounded the squeeze. First, the number of AI assistants splitting user queries kept growing — Claude, Gemini, Copilot and others were no longer a rounding error next to ChatGPT.
Second, some of the drop was self-inflicted. Sites that tightened bot protection or blocked GPTBot through robots.txt simply became invisible to ChatGPT, regardless of how good their content was.
By late 2025, the conclusion most marketers reached was simple: ChatGPT was becoming a closed loop, and referral traffic was a shrinking, unreliable channel.
Act Two: The Reversal Nobody Predicted (May 2026)
Then the pattern moved again — in the opposite direction.
On May 7, 2026, ChatGPT changed how it displays brand mentions. Previously, when your brand was named inside an answer, it appeared in bold text with no link — the user had to leave and search for you separately. After the update, that same in-text brand mention became a clickable hyperlink, pointing straight to the brand’s site.
This is the same dynamic as blue links on a Google results page: when a mention becomes clickable, people click it far more. Independent analytics tracking across thousands of sites found referral traffic rose roughly 60% per site in May 2026, and Similarweb’s own citation data shows the share of ChatGPT answers containing any citation had already been climbing — from just 0.6% in January 2025 to 2.8% by August 2025, before the May update accelerated that trend further.
Illustrative trend line, not to scale — see sources below for exact figures.
None of this reversed the citation concentration — Wikipedia and Reddit still lead — but it proves the traffic-side dial can move sharply in either direction, on OpenAI’s own timeline, without warning.
The Lesson Isn’t “Traffic Fell” or “Traffic Rose.” It’s Volatility Itself.
If you built a strategy around the July 2025 story, you’d have spent a year bracing for a channel that quietly grew 60% eight months later. If you’d ignored the whole thing as “just publishers,” you’d have missed both the risk and the upside.
The actual takeaway: OpenAI can turn this dial without notice, in either direction, and no single data point — this year’s or last year’s — should anchor your strategy. What should anchor it is durable visibility: being the kind of source ChatGPT trusts enough to name, whether it’s naming fewer brands or more.
What This Means, Category by Category
Whichever direction the dial swings next, here’s what staying visible actually requires, depending on which side of the internet you sit on.
Citation share is still concentrated in Wikipedia and Reddit. Even with clickable links now live, you first have to be one of the few names chosen.
A brand mention with no link used to be a dead end; now it can be a direct path to your product page. Being named matters more than ever, in both directions.
Feature and pricing comparisons still often get answered without a docs visit. But if you are named, that mention can now convert directly — so being cited accurately matters more, not less.
Mutual fund, loan and insurance research still often resolves in-chat. The clickable-mention update means a well-earned brand name in that answer can now drive a direct app visit.
Model and pricing comparisons get resolved inside the answer either way. The traffic swing shows why you can’t wait for a channel to “settle” before investing in citation-worthy content.
Being the cited, trusted source on a medical topic matters regardless of whether that mention is currently clickable. Trust is the constant; the link format is not.
Ingredient and routine pages need to be crawlable and citation-worthy first. Whether that citation converts on any given month depends on a dial you don’t control.
Compatibility and prediction queries can resolve fully in-chat or send a clickable mention straight to your app — the only constant is that you need to be visible enough to be named at all.
Traffic will keep swinging. Visibility — being the source ChatGPT trusts enough to name — is what survives every swing.
So what do you actually do about it?
Don’t build your strategy around this month’s traffic number in either direction. Check three things instead: whether GPTBot and OAI SearchBot can actually crawl your site, whether your content answers questions in a format AI can lift and cite cleanly, and whether you’re building the kind of authority — reviews, mentions, structured data — that gets you into the small circle of sources ChatGPT trusts enough to show, whatever the current click-through rules happen to be. That’s the AI visibility playbook, and it’s a different discipline from traditional SEO. More on how to build it, in the next post.
Sources: Josh Blyskal / Profound (July 2025 referral decline, citation share data); Similarweb Gen AI Stats 2026 (citation rate growth, May 2026 clickable-link rollout); Flavio Longato / Adobe LLM Optimizer team (May 2026 per-site referral increase analysis); SEOClarity (2026 zero-citation rate tracking). Figures are third-party estimates and may vary by measurement methodology.
Discover more from Rudra Kasturi
Subscribe to get the latest posts sent to your email.