Notice I didn’t ask “if” newer Artificial Intelligence tools like Generative Pre-trained Transformers (GPTs) will impact Google. I asked “How badly.” While Google isn’t going anywhere, these new tools will chip away at its search market share and perhaps overall value. Google and parent company Alphabet develop properties from mobile operating systems to devices like Nest and Fitbit, Waymo autonomous driving, cloud services, and more. Still, advertising from search, YouTube and their network was almost 80% of their revenues in 2023. Google’s dominance remains and its advertising revenue is still its golden goose. However, GenAI tools introduce a serious competitive threat to the core search business. How badly will Google feel the impact?
Here’s a high level summary of the main points, after which I’ll try to defend each.
- Basic Search Needs: AI GPTs are increasingly satisfying basic search needs. While not perfect, they meet many use cases where users are seeking answers; not links to maybe answers.
- AI Stickiness: GPTs offer capabilities beyond search, encouraging users to stay in that space. Even with some of the challenges with AI accuracy, people may stick with a “good enough” solution.
- GPTs are Improving: These tools are rapidly evolving, fueled by intense investment and innovation.
- Google Is Somewhat Stuck: Google’s brand is so tied to traditional search that pivoting may alienate users or undermine its core business.
- Everyone is Attacking from Multiple Vectors: Yes. OpenAI’s ChatGPT may have been the alarm bell, but there’s a whole lot more coming; both consumer and business.
- Business Use Cases beyond advertising: People are paying premium prices.
That’s the tl;dr. Stop here. Or if you want the backup rationale behind these points, continue…
Basic Search Needs
Generative AI will chip at search marketshare at an increasing rate over the course of 2025 and beyond, even though there are differing use cases for these tools. Not long ago, I’d thought that the value of traditional search, (its “truthiness” and referencing), might keep it as a powerful category on its own. But it seems clearer that adding search into AI is going to be easier than going the other way. It’s not that Google lacks the skills, (Google invented TensorFlow after all), but brand and experience lock in, and the speed with which everyone is deploying new things appears to have caught them by surprise.
There are three general use cases for search.
- Commercial & Transactions – Finding product info or transaction opportunities.
- Navigation – Seeking something specific, maybe a repeat destination.
- Information – Searching for references, often as groundwork for creating something new.
Among these, information searches are most vulnerable to GPTs. Unlike traditional search on keywords, GPTs respond to open-ended prompts with context-aware answers. Those experienced with Google likely have learned to iterate on keywords and phrases. Natural language with GPTs might actually be more typing, but the provided context may yield better results. And these results encourage iteration, which happens within context. Users can travel down an information scent path. (And this is just the use case for search; we’re not looking at other capabilities just yet.) While commercial/transactional searches and some other specific needs likely remain Google’s revenue stronghold, information use cases are under threat. GPTs streamline user experience, requiring less effort to extract meaningful results. And when we talk traditional search, we mostly mean Google. Leaving aside the 3.5% of marketshare of Bing, and less for DuckDuckGo, (which is Bing in the background anyway), and Brave search, it’s Google that’s most at risk.
This kind of threat is new to Google. We’ve seen industry shift movies before and how executives ignored the warnings. Digital photography decimated Kodak even though the company invented some of the tech. They stuck with their own agenda and lost. (See: (How Kodak Failed – Forbes.) Look at 244 year old Encyclopedia Britannica vs. Wikipedia. No contest. (See: After 244 Years, Encyclopaedia Britannica Stops the Presses.) Blockbuster vs. Netflix. (See: New York Times, Netflix vs. Blockbuster.) In this case Google seems to understand they face a serious, though perhaps not wholly existential threat.
We’ve essentially had a “winner-take-most” market with search. Now, even though Google built some early AI/GPT, how and where might they deploy it and monetize it effectively? Will doing so be a tipping point where content publishers say, “We rolled over for you in the 90s, but you wounded us. Now this?”
Meanwhile, back at the screen… consumers and businesses are enjoying the benefits of GenAI, regardless of some challenges it may suffer. Are AI results a little biased? Or just wrong? Oh well, it’s mostly ok. Besides, they’re getting better fast. We don’t care that much because we kind of got an answer at speed and being the click happy monsters we are, we’re already on to the next thing.
AI Stickiness
Let’s define satisficing. It’s when we have something that allows us to find a solution or complete a task that is “good enough” rather than optimal. It recognizes we often settle for the first satisfactory option instead of exploring all possibilities, prioritizing ease and speed over perfection. This is why we click the first search result that looks relevant, choose the first product with decent reviews on an e-commerce site, etc. It’s not laziness. We’re conserving energy in our search efforts. (For an information science perspective, see the 1989 paper from Marcia Bates, “The Design of Browsing and Berrypicking Techniques for the Online Search Interface“)
This partially explains while we’ll be spending more time in GPTs over search as our default starting place. Generative AI tools obviously expand use cases beyond search, including:
- Complex question answering and language processing: GPTs can provide detailed explanations for topics, going beyond simple keyword matching, and they can often translate languages as well.
- Context-aware responses: GPTs can maintain context through a conversation, allowing for follow-up questions and nuanced interactions. (A few years ago, I was looking for a barrel hinge to replace a broken table leaf. I didn’t know what to call it and my keywords came up short. I ended up posting a picture on a woodworking forum to get answers. I recently tested out a GPT search for this and I got right to my answer from context clues alone. No subject matter familiarity with keywords required. And that’s just text; not even an image recognition upload needed.)
- Creative Tasks: Text completion, review summarization, graphic, video, audio generation. And increasingly other content and object types from computer code to 3D models, website wireframes, blueprints; the list will go on.
So I’m sitting in this tool, my information cockpit… why would I leave to go back to some dusty ole’ keyword search? I wouldn’t. Maybe we trust Google and we’ll use their new products once they can do what we’re doing with these new things. But maybe we’re already comfortable with the new thing. Google left a window open for a first time trial of something else. And we like it. And we haven’t had something new in awhile. Even functional products can generate hedonic responses in humans. (See: Consumers Emotional Responses to Functional and Hedonic Products: A Neuroscience Research.)
The GPTs are Getting Better
In late 2024, you can’t throw a stick at the internet investment landscape without hitting something with AI in it. And while there’s increasing scrutiny on Return on Investment (ROI), this is a good thing. Because it shows maybe folks have learned a bit from the dot com bubble, the Metaverse flash in the pan, and yes, even blockchain / crypto. In other words, yes, there’s another gold rush going on, but even with the flood of money headed to AI, it’s perhaps being deployed more sensibly than in past tech evolutions.
This influx of activation energy is pushing everything. From infrastructure to code bases to tooling, etc. Yes, there are some limiting factors. Some limiting factors may be talent, but we also have chips and energy. You thought blockchain and crypto like to suck down some juice? Try AI. (See: AI is poised to drive 160% increase in data center power demand, AI Power Consumption: Rapidly Becoming Mission-Critical.) Crypto may still be using more energy, but AI is catching up fast. (See AI vs Bitcoin mining: Which consumes more energy?) Still, there’s progress in all of these areas. One of them is use of DePin (Decentralized Physical Infrastructure). Kate Laurence’s take on this in the DePin section of her Three Predictions for 2025.
These products are getting better along multiple tracks. This is partly use case specific, but also from growing Machine Learning Ops (MLOps) expertise. Especially in data prep and fine tuning. Data handling has been a segment of the tech industry forever, but now we have a whole cottage industry that’s AI-specific. And this includes “humans in the loop” tools to help with fine tuning and prompt engineering.
Google Is Somewhat Stuck
Google faces a paradox: its brand and success are built on delivering traditional search results, yet the rise of GPTs demands bold innovation. The company risks:
- Diluting its Core Brand: Pivoting aggressively could confuse or alienate loyal users.
- Cannibalizing its Revenue: GPT-style tools threaten its ad-supported business model.
Google’s success came from incremental improvements, opting for updates through extensive testing. But to compete with GPTs, it may need more radical changes, which could disrupt its own ecosystem. Even if Google develops comparable AI solutions, it must integrate without undermining its core. It’s interesting that they’ve released Gemini as a standalone product. But core search – so far – seems to only be visibly using “AI Overviews” for some searches and perhaps for how they present some results. (See Generative AI in Search: Let Google do the searching for you.) It’s understandable how they have to be somewhat cautious in integrating GPT technology, even though they’ve invented a great deal of it. YouTube may have started out with rampant copyright infringement, and even venerable Microsoft launched the Bing AI Chatbot with some results ranging from incorrect to outright disturbing. So while Google doesn’t quite face a classic Innovator’s Dilemma type problem, (because they ‘get it’), they’re still going to tread carefully as they perhaps have more to lose than the upstarts. It’s not like they can’t still innovate. Look at their Notebook Research Assistat. But will any of these things match the cash cow that their search products have been?
Everyone is Attacking from Multiple Vectors
Google’s better mousetrap once deposed early search also rans. Microsoft was generous enough to strategically keep Bing as an industry counterbalance. They maybe still make money off Bing even with its small market share, but it’s 4% vs 80%. Still, it’s kind of nice to know it’s there. (See Search Engine Market Share 2023-2024, US Search Engine Market Share (2024).
Now the GPTs are going after everything. They cover use cases Google doesn’t. What does Google have that they don’t? One thing is near real-time information. That’s huge. But changing. Then there’s mapping. For now, they likely have the most extensive language support. Plus experience understanding fraud/spam detection. And maybe other tricks. They also have cloud services and other tools. But returning to focus on core search and adjacent use cases, they face threats from multiple vectors.
Most of the GPTs give direct answers without displaying ads, threatening Google’s primary revenue stream. As more companies and consumers use task-specific GPTs, there may be less need for search. People don’t search for links; they’re looking for answers. There may be subtle background risk as well in terms of talent. Google may pay well for top talent, but some might not want to work for a huge company. They may be getting paid fine at a well funded startup and have potentially valuable stock options.
Business Use Cases Beyond Advertising
One strategic advantage Google has is its advertising system. Perhaps more so, the client base. Remember that Google (including YouTube), Meta (Facebook), and Amazon suck up 50% of global digital ad revenue. Everyone else splits the rest. (See The Rise of Google, Meta, Amazon, and Youtube in Advertising, 36 Google Ads Statistics You Should Know.)
It’s unlikely Google execs would say, “I’m glad we finally have threats to our cash cow.” But the reality is the Attack of the AIs might have come just in time to give legitimate arguments as to why Google shouldn’t be busted up for anti-trust reasons. Their revenue does seem to have slipped a bit lately. Google’s Market Share Slips, Amazon’s Rises.
Regarding advertising, an interesting point about early AI launches, like ChatGPT and Perplexity, is they didn’t immediately use ads as a revenue source. This was probably smart strategically. There’s only a handful of places to go for serious ad volume on the demand side. Yes, there’s an industry serving the 50% of the market that’s not Google/Meta/Amazon, but you’d have to face all manner of integration issues. Then there’s a digital ad industry fraught with fraud and privacy issues it seems unwilling to solve. Why start with that? Even Google’s standalone GPT, Gemini, seems to be following this paid service path. Google is incredibly mature to potentially cannibalize its own search. But this perhaps emphasizes the point… if GPTs do improve to where they take over from Search, the Great and Powerful Google is “just another GPT.” They’re back to square one in terms of competitive position.
Eventually, I think we’ll see some ad supported options. But the idea that these folks have been coming out with freemium models and not all that inexpensive paid models is very interesting. Will consumers pay Google though? People may pay for YouTube premium, but for Google Search Premium?
Back to the Bottom Line
Google has massive investment capital and talent. Large company bureaucracy notwithstanding, this elephant can pivot and move. And they’re a player in the new game. TensorFlow, Gemini, etc. It’s not like they’re on the sidelines. And arguably, they’ve been first with some of this technology, just not to market with the latest bright shiny new things.
But they’re going to take a hit. It already seems to be happening. The question will be how badly. Here’s part of your answer. How has your own information seeking behavior changed? Asking you this as a survey of one is anecdotal, not statistical, but these things still matter. Are you reaching for a GPT more often for a search need? Have you actually paid for a premium version of one or more GPTs? Even if not, for what percentage of your information seeking needs have you abandoned Google? Is it 20% or 30% or more? You’re probably not alone. (Google may have a ChatGPT problem, Number of ChatGPT Users (Dec 2024))
If you liked this article, here’s the beginning of a three part series on how Search may change with the rise of GPTs. Search Tools in a GPT World.