The zero-click story was always wrong. ChatGPT just made it obvious.

OpenAI now has an economic reason to send you to a website, and on 7 May it started doing so. The "zero-click AI kills the web" thesis assumed the platform had no incentive to drive clicks. That assumption was never sound, and the architectural reversal is now visible in the product. What follows from clickable brand names is not better UX. It is a monetisation surface.

The feature, and what it actually is. On 7 May 2026, ChatGPT began turning brand-name mentions inside answers into clickable hyperlinks pointing at brand homepages. There was no announcement. The change was discovered and quantified by Profound, an AEO (Answer Engine Optimisation — optimising for placement in AI-generated answers rather than search results) monitoring firm, which reported the rollout as "the largest single-day change in AI brand traffic measured all year" and claimed OpenAI referral traffic to its monitored brand panel roughly doubled overnight.¹

Treat the magnitude with care. Profound sells a product (Noble Nodes) whose entire premise is that brands must invest in managing their presence inside AI answers. A firm with that book of business reporting a dramatic traffic event on the day a new traffic-driving feature ships is not an independent audit. The directional finding, a new mechanism exists, traffic moves through it, is almost certainly correct. The "doubled overnight" headline is the upper bound a motivated source would produce from an undisclosed panel.

The interesting question is not the number. It is what the feature tells us about where OpenAI is going.

The zero-click consensus rested on a bad assumption. For roughly eighteen months, the dominant framing in marketing and publishing has been that conversational AI kills referral traffic by answering questions inside the chat surface. Users get the answer; the website never gets the visit. There is real evidence behind this for general informational queries, and publishers have lost meaningful organic traffic to Google's AI Mode and to ChatGPT's earlier no-link answer style.

What the zero-click consensus missed is that a platform absorbing the inference costs (the cost of running the model in production, separate from training) ChatGPT runs at cannot, in the long run, treat outbound traffic as a waste product. Outbound clicks are the substrate of every working internet business model. Affiliate revenue runs on them. Referral deals run on them. Sponsored placement runs on them. Brand-trust metrics that justify enterprise contracts run on the assumption that users can verify what the model said by visiting the source.

OpenAI did not ship clickable brand links because users were complaining. It shipped them because the product needed an outbound-traffic surface before any of those revenue lines could exist.

Google 2002, compressed. Google's path from organic blue links to a dominant paid-search business took the better part of a decade. AdWords launched in 2000, but the transition from "search engine with ads" to "ad business with a search engine" was gradual enough that most users did not notice the inversion until well into the 2010s.

The same architectural sequence is now visible in ChatGPT, except the gap between steps is months rather than years. Brand mentions became brand links in a single feature drop. The next step, distinguishing organic brand links from paid brand placement, is a UI decision, not a research problem. OpenAI's commercial team understands this. So does anyone who has read Cory Doctorow on platform decay: the useful version of the feature is the wedge that makes the extractive version shippable later.

The useful version of the feature is always the wedge that makes the extractive version shippable later. That is not cynicism. It is the only business model the cost structure supports.

I would not bet on sponsored brand links arriving in 2026. I would bet on them arriving before the end of 2027, and I would bet harder that the first version will be labelled in a way most users do not register.

The second platform shock for marketing teams. Distribution strategy teams spent 2025 rewriting their playbooks for Google's AI Mode. They are now being asked to rewrite them again, for a second answer surface with different ranking logic, different link policies, and a different monetisation roadmap. Search distribution is no longer a Google monopoly under threat. It is a Google–OpenAI duopoly with Perplexity and Gemini as smaller nodes, and each platform's "how do we get cited" problem is structurally distinct.

The practical implication for any brand that depends on inbound digital traffic: AEO is no longer a curiosity adjacent to SEO. It is a parallel discipline, and the platforms it targets have explicit commercial incentives to make placement scarce. The free phase is the current phase. It is also, on the Google analogue, the shortest phase.

Where the analogy might break. I want to take the counter-case seriously, because the Google parallel is doing a lot of work in the argument above and parallels of this kind tend to over-fit.

Two things might genuinely be different. First, paid placement inside a conversational answer carries trust risk that paid placement on a results page does not. A user who sees a sponsored link above ten organic links has a clear mental model of the trade. A user who is told, in flowing prose, that the best option for their problem is Brand X, without realising Brand X paid for the mention, has been deceived in a way regulators will notice. The FTC's existing rules on undisclosed endorsement apply cleanly here, and OpenAI is large enough to attract enforcement attention.

Second, the panel Profound is measuring is undisclosed. If it skews heavily to large consumer brands with strong existing recognition, the "doubled overnight" figure may not generalise to the long tail. A B2B SaaS vendor with three brand mentions per week in ChatGPT is not getting a doubling of anything meaningful.

Neither of these breaks the structural argument. They affect timing and magnitude, not direction. The direction is clear: OpenAI has built the rails for an advertising-adjacent business, and the rails were built before the announcement, which is the order these things usually happen in.

Footnotes

¶