How Google Profits from AI-Powered Search, While Killing SEO
Making SEO harder is more profitable for Google
As announced in Google I/O, Google will roll out Search Generative Experience (SGE) to everyone in the U.S. soon, and to all international properties sometime later this year.
I wrote about how this move will “flood” the consumer market with GenAI, and crowd out the top-of-the-funnel of GenAI to make OpenAI’s life difficult. After all, only a small fraction of Google users (which is in billions) have yet to encounter ChatGPT yet. Google will be the first GenAI experience for many international Google users.
But there are other reasons why putting GenAI into search will be great for Google, especially Google Cloud:
It immediately gives the biggest “operating leverage” for GCP in terms of generative AI, since Google search has the largest user base.
Google will have the highest GPU utilization % due to dogfooding its own GenAI infrastructure, which Google has vertically integrated with great success (with its TPUs)
.. which in turn might give Google some long term “edge” in competing with Nvidia, etc, in commercializing its chips, which will take some years to shape up.
Not only that, its entire Gemini and VertexAI product suite will get dogfooded by the world’s largest GenAI customer (Google Search), giving a compelling story for GCP’s sales teams selling VertexAI and peripheral products.
Also, while SGE (GenAI search) will absolutely make small publisher’s lives hell, it will ironically boost Google’s revenues, at least in the short term. Here’s how.
With AI overviews, there’s even fewer “slots” for links to be seen by consumers, especially if the AI overview actually works well. The best indicator of AI overview’s utility would be the user not checking out any link, at all.
To compete for these fewer “slots”, advertisers might pay even more money to make up for declining traffic, or be forced to bid on keywords to get discoverability.
The roll out of “multi-modal queries” (e.g. circle to search, video search, etc) will discoverability even worse. For example, Alexa tried to build an ad business but failed miserably because it’s extremely awkward to put sponsored content in any medium other than text.
Losers will be small publishers. “great content” may still rank and get incorporated into AI overviews, and it won’t stop people from creating content, but things are looking dire.
Of course, there are still plenty of ways Google can mess up. Putting GenAI into search will not come without occasional setbacks, because the UX is still being figured out. Perplexity and ChatGPT have the luxury to keep the UI simple, because it mainly serves research queries. But Google search needs to work for a whole bunch of other queries, such as:
local search, business search
navigational queries
commercial queries
location queries, maps, etc.
Thus, I expect Google to:
hedge its bets by not rolling out AI overviews indiscriminately, especially for the most profitable keywords that account for 80%+ of Ad revenue (Insurance, Finance, Mortgage, etc).
prioritize GenAI deployment for 1) long tail keywords, and 2) research intent queries, which overlap significantly with Perplexity and ChatGPT
Side note: will this kill Perplexity? I don’t think so - I think if anything being exposed to Google’s GenAI answers might become “free advertising” for apps like Perplexity, which aims to keep things simple. Perplexity - for now - is a superior experience to Google’s SGE in many ways:
It has low latency. It cites good sources. It generates a succinct and not overly dense solution. It reduces cognitive load with pre-generating follow up questions. It doesn’t require a sign in, etc.
That being said, given that Perplexity is fundamentally a LLM wrapper (while being a good one) - its UX elements are not an insurmountable moat, and nothing’s stopping Google from publishing a more simpler, researcher-friendly version of Google.
And clearly, OpenAI will be offering GPT-4 level ChatGPT mobile app for free, so I’d be more concerned about OpenAI if I were Perplexity founders.