As The Economist noted last week ("AI is killing the web. Can anything save it"), the rise of large language model (LLM) chatbots is transforming how people find information online, whether they’re looking for the most stylish sneakers or the best treatment for hypertension. But this shift isn’t just a challenge for search engines. It threatens the entire ecosystem of companies that, over the past two decades, have built their business models around internet search, relying on it to attract visitors, sell products, and convert clicks into revenue.
Now when you Google a product or service, or worse, ask ChatGPT, instead of getting a page of links, many of them sponsored, you get a neatly packaged summary, pulled from what the AI “knows” or finds though its own searches in the background. That means fewer clicks, fewer ad impressions, and ultimately less revenue for companies dependent on search-driven discovery. Ad revenues are falling. Click-through purchases, if they haven’t declined already, soon will.
Where does this all lead? One clear implication is that companies which previously advertised on Google—I use Google here as a stand-in for search generally, but with nearly 90% of global search ad revenue, it's effectively the platform—will need to find new ways to reach audiences. Firms that once paid to appear in search results may instead start blocking AI scraping or begin charging them for access. Rather than relying on search-based advertising, they may pivot toward more targeted outreach through social media, and place more content behind paywalls.
This shift also brings back a debate around proprietary information and copyright. AIs like Google’s Gemini and OpenAI’s ChatGPT generate responses by aggregating and summarizing content created by others but without compensating the original authors. As more of the web moves behind paywalls or becomes inaccessible to web crawlers, AI “information aggregators” will only be able to draw from what is left in the public domain. The result will be a narrower and potentially distorted AI knowledge base, increasingly shaped by what’s freely available rather than what’s most accurate, insightful, or commercially valuable.
One long-discussed but still unrealized solution would be to embed copyright metadata into online content and implement an automated micropayment system. Under this model, AIs would compensate content creators whenever their work contributes to a response. Though the idea has been discussed for years yet never operationalized, the rise of generative AI may finally create the incentive to bring it to fruition. Perhaps we’ve now reached the tipping point where that vision becomes reality.

