AI Isn’t Killing Search. It’s Expanding the Search Pie

“ChatGPT traffic analysis: Insights from 17 months of clickstream data”* *Source: [Semrush](https://www.semrush.com/blog/chatgpt-search-insights/) For the last year, the lazy take has been that ChatGPT is going to replace Google, but ChatGPT is not replacing search. It is stretching it. People are not just moving from one search box to another. They are building more complex discovery journeys. They ask ChatGPT for advice. They check Google. They click through to websites. They compare sources. They bounce between AI tools. They come back and refine. The old journey was. Search → click → website → purchase The new way pf doing things is. Prompt → answer → shortlist → search → compare → click → purchase And in that journey, brand visibility inside AI answers becomes commercially important. The headline: ChatGPT is becoming a referrer The Semrush analysis found that outbound referral traffic from ChatGPT grew 206% in 2025. That is the important bit. Not because ChatGPT is suddenly sending Google-scale traffic. It is not. But because the direction of travel is clear. ChatGPT is not just a closed answer engine. It is becoming a discovery layer that sends people out to the wider internet. That makes AI visibility measurable in a more practical way. If people discover a product or brand in ChatGPT and then click out, search again or visit directly, AI search is now part of the acquisition journey. The problem is that most brands are still not set up to see it. Google still matters! One of the most interesting findings was that over 20% of ChatGPT referral traffic goes to Google. That should kill the “SEO is dead” narrative. AI search and traditional search are not separate worlds. They are connected. People use ChatGPT to frame the question and Google to validate, navigate or continue the journey. For brands, this means the old fundamentals still matter: - Strong category content. - Clear product pages. - Authoritative off-site mentions. - Structured data. - Reviews. - Good technical SEO! But those foundations now need to work for two audiences: humans and machines. Your content needs to help a person make a decision. It also needs to help an AI system understand who you are, what you sell, what you are good for and when to recommend you. Most ChatGPT prompts do not look like SEO keywords This is the shift that gets missed. According to Semrush, for much of the study period, 65% to 85% of ChatGPT prompts could not be matched to traditional search keywords. That makes sense. People do not talk to ChatGPT like they talk to Google. They do not just type: “best running shoes” They ask something closer to: “I’m training for my first marathon, I overpronate slightly and I want something comfortable for long runs that won’t feel too heavy. What should I look at?” This is why brands need to move beyond keyword lists and start thinking in terms of prompts, problems and subjective product needs. What is the customer trying to solve? What trade-offs are they weighing up? What language do they use when they describe the situation? What makes your product the right answer? That is where AI search optimisation starts to become useful. ChatGPT only searches the web some of the time! Semrush found that ChatGPT enabled search on roughly one-third of queries as of February 2026. This is a big point. A lot of ChatGPT answers still come from the model’s existing knowledge, not live web search. That means brands need to think about two types of visibility. The first is live retrieval visibility. This is when the model searches the web and finds current sources. The second is model knowledge visibility. This is when the model already “knows” enough about a brand, product, category or source to include it in an answer. For live retrieval, you need strong, crawlable, well-structured content. For model knowledge, you need consistent public signals across the web. That includes your own site, retailer pages, marketplaces, reviews, press, forums, YouTube, Reddit and other sources models may learn from or retrieve. This is why AI search is not just an SEO problem. It is a digital shelf problem and the new shelf is invisible... On Amazon, brands understand the shelf. If you are not ranking, you are not seen. If your PDP is weak, conversion suffers. If your reviews are poor, shoppers hesitate. AI search creates a similar shelf, but it is harder to see. When someone asks ChatGPT, Gemini or Rufus for recommendations, the answer is effectively a shortlist. If your brand is not included, you are not on the shelf. That is the invisible shelf. And unlike Google, where you can see ten blue links and measure position clearly, AI answers are more fluid. They vary by prompt, platform, context and user intent. So brands need a new operating rhythm. Track the prompts that matter. Measure whether you are mentioned. Measure where you appear. Look at the sources being cited. Improve the content and authority signals that influence those answers. And repeat. The final point. ChatGPT referral traffic growing matters. But the click is only one signal. A lot of the value of AI search will happen before the click. It will shape awareness, trust and consideration. It will decide which brands make the shortlist and which brands never enter the conversation. Search used to be about ranking on a results page. AI search is about being recommended in the answer. For eCommerce brands, that is the next battleground.