Google is in a predicament. It has a hugely profitable business built on traditional search. Now external pressure is forcing Google to roll out AI search. The problem is that AI search means fewer clicks, fewer websites which can survive to power search and ultimately less money for Google. At first glance, that looks like Google’s problem to solve. But the structural pressures reshaping how Google operates search are the same ones reshaping how you should be buying it.
The trap
For nearly 30 years, Google and the open web functioned on a simple exchange. Google crawled the web and returned traffic to content owners. Google got to own the search space and monetise a subset of queries, while the website owner had the traffic to drive advertising revenue, sell products, and make more content. Google needed that content. It’s what powered search and what kept people coming back.
AI search breaks the exchange. Searches without AI Overviews already have a 34% zero-click rate. With AI Overviews, that rises to 43%, and in full AI mode it’s 93%. Less traffic means less ad revenue for publishers, which means less content, which means worse search. The thing Google needs to power search is being eroded by AI search.
Publishers are acutely aware of their eroded position. Cloudflare, which handles traffic for roughly 20% of the web, now blocks AI crawlers by default and is piloting a pay-per-crawl model. The crawl-to-referral ratios tell the story: Google’s is 14:1, OpenAI’s is 1,700:1, Anthropic’s is 73,000:1.
For advertisers, the implication runs parallel. As third-party publisher inventory thins and zero-click rates rise, brands that produce genuinely useful, well-structured content have a structural advantage. Not just for traditional SEO rankings, but because AI engines cite sources when they answer queries. If your content isn’t being cited, you’re not in the answer.
The monetisation model is changing shape
Google’s current ad model is unusually elegant: a self-optimising live auction where advertisers compete for clicks, quality scores maintain relevance, and Google takes margin without having to curate the content. AI search will involve fewer clicks, so a product comparison that previously took three clicks now appears in a single AI answer. Fewer clicks means fewer auctions, which means less revenue per query.
What replaces it isn’t settled, but the options are taking shape. Sponsored placement within AI responses, cost-per-action models where advertisers pay when AI completes a transaction, sponsored context where a brand pays for the recommendation itself. Ads inside Gemini itself are openly speculated about, denied by Google publicly, and almost certainly being planned in some form. And if Google does go there, they have 20 years of ad infrastructure and data advantages that Perplexity and ChatGPT simply don’t. Whatever format emerges, Google is probably better placed than anyone to make it work.
The question for advertisers isn’t whether Google figures this out. It’s that none of these formats exist in the form advertisers know how to operate. The pricing and auction dynamics you’ve spent years optimising don’t yet exist in their new form. The strategic question now is how to build familiarity with emerging formats before those models settle and the window to move early closes.
The product war is the part that matters most for advertisers
Even if Google solves the content supply problem and figures out AI monetisation, it only matters if users keep coming. That habit is no longer guaranteed.
OpenAI, Perplexity, Anthropic et al are all trying to be better. Unlike Google they have no revenue conflict in doing so – they can optimise purely for the user at a fast pace. Google has to weigh every product decision against its impact on the existing business. That makes them structurally slower in a moment that rewards speed.
The bitter irony of course is that Google knows all about this trade off of better answers vs existing revenues. The transformer architecture that powers ChatGPT, Gemini, Claude and every other large language model came out of Google afterall. They didn’t miss the AI wave. They built it. They just couldn’t be first to ride it in search without undermining their existing revenue streams.
The bottom line
Google is not going away, and it will find a version of this that works. But the transition is real and already affecting how search functions. Google is caught between moving fast enough to stay relevant and slowly enough to protect the revenue that funds the transition. That tension won’t resolve quickly, which means the search landscape you’re planning against today looks materially different from the one you’ll be operating in three years. The question worth sitting with now is which parts of your search strategy are built on assumptions that no longer hold.
