The short answer seems to be “Yes.”
This isn’t really a new idea at this point. But I wanted to offer my take on it as I don’t think I’ve seen a good overall summary and I wanted to offer one.
The deeper questions seem to be: How will it hurt? How much? And what will Search companies and Publishers do in response? While Generative AI is different from Traditional Search and has different use cases, some aspects are taking market share from traditional search. According to Similarweb back in 2023 “ChatGPT Grew Another 55.8% in March, Overtaking Bing.” And there’s the early 2024 prediction by Gartner of Search volume dropping 25% by 2026. Then AdWeek said that Google’s use of AI alone could hit publishers with a $2B annual ad revenue loss.
This will likely continue and is somewhat ironic. Search will take market share hits because someone else is using content that isn’t theirs to create value. Sound familiar? Seems like the foot is on the other shoe. And yet, while Search might be facing some disruption, their core value propositions remain just fine. Also, they typically are integrated with several products; e.g., Google across web, video, maps, and more. The real question is how they’ll deploy billions to respond, not that anything’s replacing them soon. Still, with hundreds of billions at stake, even a small percentage hit is non-trivial. (Even if Google is growing ad revenue faster than others lately.) We should be more concerned with the smaller search engines and the publishers themselves.
Web Search has made its riches from publishers’ content at the same time as it’s somewhat disintermediated them. Google alone made over $200B in 2023. That’s a good chunk of the industry’s revenue. Publishers could do little about this as the alternative, (to not be in Google), is untenable. Now Generative AI (GenAI) – from ChatGPT to Claude, Perplexity and more – is taking market share from search for some information needs use cases. Unlike search, GenAI tools mostly don’t show content references. Of course, the majors have their own efforts, like Google’s Gemini, and Microsoft is trying to fly copilot, but these may have to compete on their own merits, not just as embedded products. Incumbents wield power with their moats, but can stumble or get hacked apart one small vertical market at a time. They need to be smart. If Microsoft tries to just stuff copilot into everything like it was Clippy… well… that would be unfortunate.
Publishers have endured a tough digital transformation, earning less as the so-called “print dollars” turned into “digital pennies,” with Cost Per Thousand (CPM) rates dropping from $15–$30 in the 1980s to $1–$10 today. Now, as GenAI tools generate content summaries, traditional publishers may fall further down the value chain. Attribution remains a problem—users receive answers without knowing whose content informed them, and even if GenAI models attempted to link back, accurately tracing content sources would be costly and technically complex.
This article isn’t about GenAI tech, but suffice it to say, the data training process and encoding is not amenable to reference without significant work to make that happen. If I asked you, “What’s the moral of the story from ‘The Tortoise and the Hare’?” you might give me a good answer. But if I asked, “Can you give me a reference for that?” What then? Maybe you’d remember it’s from Aesop’s Fables. But is that it? Maybe you learned from a parent or a teacher? More likely all of them. Your knowledge is mixed together. If I had you attribute your knowledge and apportion the percentage of it to each source, could you do it? Unlikely. But let’s say you could. Now, what if you have to split apart any subscription or ad revenue you earned on that answer. (Interestingly, GenAI seems to be trying for paid subscription revenue. I wonder how much howling we’d hear if they started slapping ads on as well. Maybe we’ll find out.) The reality is attribution for both sources and revenue is hard. Right now, some GenAI engines seem to be at least attempting to get better source attribution. This will be challenging. Data architecture and encodings would need to change before foundational model training. This will be expensive and hard to evaluate. And my example was simple text, not imagery, sound, video or other content types. Still, where there’s a will, there’s a startup. ProRata.ai thinks they have an attribution solution for some of this.
And you thought the music industry was challenged with figuring out and allocating streaming revenue share? What’s happening now is more mind bending. Focusing on news in particular, journalism Professor Jeff Jarvis, among others, have said something to the effect that “We don’t need newspapers, but we do need journalism.” The same is true for quality content in general, however you choose to define quality. If content producers suffer further revenue destruction, what will happen to quality? The news business has already been decimated. What happens when everything else suffers. The point is these issues are existential problems; not just legal and revenue attribution issues. We seem ill-equipped to manage the complexities of what the greater good might look like here. These are important issues, but I’ve gone a bit off the rails here… Moving on then…
Traditional Search Vs. Generative AI
Leaving aside what might be irony calling something under a few decades old “traditional,” it’s a strange market. While there were a variety of search startups during early web growth, upon Google’s ascendance, it became a winner-take-most marketplace. Microsoft runs Bing as one small counterweight to Google, but at just a few per cent of the market. Even though a distant second place, Microsoft earns billions from Bing, (over $6B in 2022), and grew nicely so far in 2024, and there’s also powerful strategic reasons to keep it going. Meanwhile, there’s some small dogs nipping for share with niche value propositions. Some are traditional also-rans like Ask.com, and others focus on things like privacy, such as Brave’s search engine.
And now comes Generative AI. It’s not search, but can substitute in many cases. And that’s the worst. Because you can compete with others in your industry. You can battle it out on a feature-by-feature basis, or try some slightly different directions. But a substitute is harder. Consider GPS companies when the iPhone showed up. Now you’re competing with a device where your entire product is one minor feature. You’re just drive by collateral damage. With that as background, let’s get back to our Search story… Even if you try to make a better hybrid product, your brand mindshare is locked. From a user experience perspective, customers have decades of muscle memory for how you work. And if you try to force all of these new features into your product maybe that just damages what you’ve got. We’re not just talking about the new marketing guy’s small branding change that annoys customers who can’t find their favorite soup because you changed the label. This is actual functionality.
It’s About the Use Cases
What kinds of things have really changed? Specifically. Let’s look at some of these things more in depth past the usual, “Well, the Generative thing makes stuff.”
TradSearch Use Cases
Following are short summaries of search use cases. Each could easily have its own book. Some do.
- Commercial & Transactions – Finding product info or transaction opportunities.
- Navigation – Seeking something specific, maybe a repeat destination.
- Information – Searching for educational or reference material, often as groundwork for creating something new.
GenAI Use Cases That Impact Search
- Content Creation (multiple media types) – GenAI can create new material, making search less relevant as a middleman to just seek source material.
- Content Discovery (search and analysis) – Users can now access summaries instead of full results. The GenAI summary snippets search engines are providing now are ok, but not iterative. It’s a good feature, but only partly helps with one use case. It’s good. But not necessarily great.
- Content Disambiguation (search and discovery) – GenAI can infer meaning from “fuzzy” queries. In search, a user may lack keywords for an idea. But describing something in human terms may offer enough context to collapse any ambiguity. This can be a non-trivial use case for those traversing new-to-them information spaces.
- Conversational AI (summaries and assistants with natural language) – Smaller information spaces, such as customer support systems, might address user questions locally, without search engines.
GenAI Use Cases that Don’t Impact Search
Many GenAI use cases—like generating new art, whole stories, or school lesson plans—don’t overlap with search. (Or rather, they “sort of” don’t.) I say they don’t overlap because traditional search just doesn’t do these things. However, as GenAI draws users for these tasks, overall search traffic will nonetheless likely decline. Because part of why users went to traditional search was for source material, which won’t always be needed now. Unlike traditional search, GenAI generates new responses tailored to users. And where search is needed, in the future it’s likely easier to embed simple search functions into GenAI than to retrofit AI into search.
Death by a Thousand Cuts
GenAI’s impact goes beyond big players; it’s embedding itself in chatbots and niche knowledge bases. This reduces the need for traditional search, as these specialized systems offer users targeted answers without search engines. In many cases, these will “small language models” with specialty knowledge bases for their particular needs. This information might not even be available in either traditional search engines or another major AI because it never even had access to this data. Look at how Amazon ended up being as much a place to do product searches as it is for shopping. Even back in 2021, NielsenIQ pointed out 6 in 10 shoppers began their product search on Amazon.
What About Content Publishers?
How bad is all of this for Publishers? The technology sea change of GenAI is one of those places where you can legitimately use the otherwise abused labels like paradigm shifting. This isn’t quite as severe as cars replacing horses or “video killed the radio star,” but there’s a big fat chunk of use cases that can easily be departing from traditional search and going to generative AI.
In terms of content publishers in particular, we have at least the following business issues:
- Reduced Traffic to Sites – Direct answers in search engine summaries may preclude the need to go to a site, even more so then with the link snippets, answer boxes or knowledge panels search engines have provided in the past.
- Aggregation without Attribution – Source usage happens without attribution or compensation.
- Erosion of Value – As GenAI content floods digital spaces, original content could get crowded out. It may be hard to justify the costs of high quality content if users go to free, summarized information.
- Lowered Revenue – Much of content is ad supported. Less traffic means shrinking budgets. And paywalls / subscription models may suffer as well.
- Copyright Enforcement – As discussed, it will be challenging to enforce intellectual property rights.
- Audience Fragmentation / Brand Dilution – If content is delivered via AI summaries, even with attribution, publishers lose control of their brand and brand voice. Information becomes generic.
In short, this is a deeply challenging issue. GenAI clearly provides a great deal of value, notwithstanding its risks and limitations. Just as clearly, there’s systemic risk of quality content value erosion. We do not yet seem to have clarity on any balanced solutions to these issues.
GenAI Can Likely Integrate Search Better than Search Can Integrate GenAI
It’s simpler to add basic search functions to GenAI tools than to integrate GenAI capabilities directly into search. While we’re seeing summaries on search search results pages, it’s a limited offering that could actually push links to publishers further down, impacting revenue. And the iterative prompt experience is just fundamentally different from getting a results list. Of course, there’s plenty of other ways Search can potentially integrate a GPT/AI experience. (Meanwhile, I drafted this article a couple of days ago. Today I got an email announcing “Introducing ChatGPT search.”)
My assertion here is that traditional search and GenAI are fundamentally different user interfaces and experiences. GenAI’s iterative, context-driven approach is intuitive, and iterative, while the search engine model is deeply ingrained, offering quick, isolated answers. Usage behaviors are hard to change. People have decades of built in expectations as to how these things work. Search gets you an answer and you’re done. GenAI is new, but we already seem to understand it’s iterative. And the interface is reasonably smooth. A search engine can’t just change this baked in mental model. But we’re likely going to see them try. Even with their reach and power though, no one is going to easily change that behavior. (This is a topic I may try to cover more in depth, but beyond scope for now.)
What to do, what to do?
I’ve made some assertions about what’s going on and what could happen next as the emergence of GenAI impacts existing search engines and publishers. As to what incumbent search engines and publishers might do? These are really separate topics deserving of their own treatment. Maybe I’ll attempt to offer some thoughts on these at some point. It’s a classic Art of War product and marketing exercise. As much incremental change as we’ve seen over the past decade or so, things have maybe been a bit stagnant in the information retrieval world. Still, with the average lifespan of companies having declined over the decades, they might start moving faster. (Statista calls it as 21 years as of 2020, and EY says just 15 years in 2023.)
So we’re seeing the old proverb still applies…
“Every morning in Africa, a gazelle wakes up, it knows it must outrun the fastest lion or it will be killed. Every morning in Africa, a lion wakes up. It knows it must run faster than the slowest gazelle, or it will starve. The lesson? It doesn’t matter whether you’re the lion or a gazelle. When the sun comes up, you’d better be running.”