I initially planned to write about how Generative AI (GenAI) might impact traditional search in a classic startup vs. incumbent scenario. Over time, as the landscape rapidly evolved, first I shared thoughts on “Search Tools in a GPT World” and then “Traditional Search vs. GPT Business Models.” This all leads to the obvious question… what are traditional search engines doing and what else might they do to respond?
What’s the Threat. Really.
Is this just about eyeballs and usage again? No. It’s not. The rise of fee-based offerings and other models changes things. (See my earlier post, “Will Generative AI Hurt Search & Publishers?” and the section “It’s About the Use Cases.”) The risk here is about shifts to products with different business models. Let’s assume GenAI poses a legitimate threat to search engine dominance. While individually they may not seem formidable, collectively they challenge the status quo. While free and freemium models (including embedded options) will persist, fee and subscription-based products are set to reshape the landscape. Issues of fairness and benefits regarding the haves and have nots is not part of scope here, but clearly this is also a potentially seismic shift in the consumer market.
Search and GenAI serve different needs. While both operate in the information space, they diverge in approach. Search retrieves and ranks existing data, while GenAI synthesizes information into probabilistic text. When queried, GenAI generates new text by summarizing or extrapolating from learned data patterns. This raises a question: is GenAI search or just content generation?
I was midway through drafting this, (weeks before I got around to finishing), when OpenAI announced “Introducing ChatGPT search” integrating live web search. Many expected some such offerings, but this development looks fairly polished for a 1.0 go at it. Google, with its history of well-tested incremental changes, faces the need to consider large UX changes to compete or risk commoditization. (Or obsolescence.) The launch of standalone Gemini highlights Google’s reactive approach trying to match competitors feature-for-feature. Gemini won some points over the summer for an update that gave faster results. That’s great. Then again, is this what Google’s latest product has come to? Just slugging it out feature-for-feature with other peers and near peers? Despite having experts who helped invent core technology, Google is now competing feature-for-feature. That’s hardly the dominance they’re used to. Maybe this will be good for them with any Department of Justice anti-trust conversations. Because this feels more like a commoditized business; not at all a monopoly.
Second Mover Advantage
Let’s consider Second Mover Advantage. Most startup watchers are familiar with “First Mover Advantage“. Not everyone is as conversant with second mover advantage. Here, follower businesses have seen the industry evolve. There’s an ecosystem, advertising technology and more. Later entrants can mix and match the best pieces, ignore others, and select strategies for attack without the extreme expenses of market creation. The old idea comes to mind; “it’s often the second mouse that gets the cheese.”
Google is partially stuck. Or at least faces major strategic choices. They have modified their results pages over the years, but their fundamental experience has famously not changed in practically forever. They’re faced with either a radical, (and disruptive to customers), change in their experience, (which will be a risk), versus an incremental response, (maybe too slow), and doing nothing, which would be devastating. Alternatively, they can introduce separate products. (Which they’re doing with Gemini.) Will this splinter their offerings? This is maybe not the worst thing ever for Alphabet writ large. But for Google in particular? It starts to move from behemoth gorilla to a niche product. Or perhaps worse, a commoditized product.
What Can Search Do in Response?
There are two high level answers. Well, maybe three; Fully Integrate, Create New Products, and Some Kind of Hybrids. Let’s take a look at these.
Fully Integrate
This approach involves recasting products with embedded AI. Competing head-on with startups signals commoditization and strips differentiation. Google’s moat could shrink to its ad system and commercial relationships. Though their cash allows for acquisitions.
Look at Perplexity. They nicely integrate search with generative AI. Others do similarly, including Gemini. There’s no history of what the best design patterns might be for integrating this yet or which combinations of features will resonate. Perplexity grew from the ground up to do this though. Google and Bing just started using AI summary boxes in search results; not a big leap. Unfortunately of course, all the Gen AI products may suffer from quality issues, hallucinations, etc. whether it’s ChatGPT, Perplexity or Google. But might Google’s brand suffer most with any credibility issues? Or will Google prevail in the end? Tim Gordon says, “ChatGPT’s threat to Google is that it makes search worse, not better” and “There is no reason why they cannot reformat their proposition. Access to Generative AI technology is not a competitive advantage against an AI behemoth like Google.”
Create new products
Standalone GenAI products allow new value propositions. Combining features like search history, project management, and collaboration could enhance offerings without compromising search. Each product can find its own value propositions.
Another standalone product might be to ramp up paid API access services to serve general AI companies. AI tools in many cases can benefit from having their agents use indexed search results as part of their prompt and RAG processes.
Hybrid Approaches
Full integration could be a bad idea. It might work, but risks degrading the core product, possibly smashing the existing business model. Not to mention potential quality issues.
Are there ways to integrate some kind of hybrid solutions that could integrate some of the best of the new and maintain the quality of the old? Maybe.
What are some features that might be adopted as useful in a hybrid approach?
- Layered Search Results – Combine search listings with AI summaries or insights alongside. Users get direct answers and the option to explore deeper sources.
- Interactive Query Refinement – Use AI to assist refining searches, allowing iterative shaping of queries without replacing the standard search interface.
- Parallel Search Modes – Offer a toggle between standard search and AI-powered exploratory modes. Users can choose depending on the complexity of their needs.
- AI Summaries with Source Links – Present AI-generated content with search results and references to sources, blending AI synthesis with verifiable information. (Some of the Gen AI solutions are doing this now as they try to integrate real time search results.)
- Project and Task Integration – Expand search to support project structures where users can save, organize, and collaborate on research and queries over time.
- AI Plugins and Extensions – Allow third-party AI models/tools to integrate with search, providing specialized insights while maintaining the traditional search backbone.
- Multimodal Search – Incorporate AI-driven voice, image, and video analysis into search while preserving classic text-based search as a complementary option.
These approaches could allow traditional search engines to leverage the strengths of GenAI while retaining familiar, trusted elements of their platforms.
Marketing
Consumers may need education as to where traditional search has advantages over GPTs. Traditional search retrieves comprehensive, exhaustive results directly from indexed databases, better ensuring no relevant data is missed. In contrast, LLMs can get things wrong, even when using augmented retrieval. This can result in incomplete or imprecise answers if the vector search part of the augmented retrieval fails. For focused use cases such as support chat ‘bots, this might work well enough, but less so for general tasks. Traditional search excels at broad, exact matching, while Retrieval Augmented Generation (RAG) enhanced LLM can shine in contextual synthesis. (Though again, risks gaps when retrieval falls short.) This makes traditional search more reliable for exhaustive queries where full coverage of a topic is critical. Traditional search also generally provides consistent, repeatable results for the same query. There’s better transparency and control. And pure search is likely better at handling precise keyword queries where exact terms matter (e.g., product SKUs, legal phrases).
This could be an uphill battle against the new. Search may offer more exhaustive results. But there’s a joke about this. Question: Do you know one of the best places to hide information on the Internet? Answer: The second page of Google Results. No one goes there.
How Generative AI Upends Search
As I’ve pointed out in a prior article, GenAI is a different kind of tool with varying use cases. This is the challenge for traditional search. While Search still does things better than GenAI, (basic navigation, transactional), GenAI is arguably better for a variety of information seeking tasks. GenAI can provide richer, contextual answers. This presents a threat to incumbent search engines whose dominance relies on structured data and precise algorithms.
For some use cases, users may prefer conversational interfaces and tailored summaries over sifting through links. This shift redefines the utility of search engines, pushing them towards either fully embracing AI-powered synthesis or risking obsolescence as users gravitate to platforms like ChatGPT or Perplexity.
What the Incumbents Have Done So Far
Incumbents like Google and Microsoft haven’t stood still. They’ve invested heavily in AI-driven features, with Microsoft embedding OpenAI technologies directly into Bing. Google’s Bard project exemplifies their commitment to integrating AI into search workflows. However, most implementations remain incremental, such as search summaries, AI-enhanced suggestions, and refined ad placements.
The challenge lies in balancing innovation with user expectations. Search leaders are cautious about disrupting the core experience that users trust, leading to slower, more measured rollouts compared to the radical AI-first approaches of newer players.
Major Risks
There may be many risks when facing disruptive new products in your market. Here’s what I believe are the top three:
- Multi-pronged Strategic Business Threats. We can turn to Michael Porter’s Five Forces and see that incumbents face issues along all five dimensions, Threat of New Entrants, Bargaining Power of Suppliers, Bargaining Power of Buyers, Threat of Substitutes, Industry Rivalry.
- Modifying your core product may risk alienating current customers.
- Significant Cost Structure Issues: As pointed out in an earlier article, “Traditional Search vs. GPT Business Models,” the cost structures of GPT technology is different than traditional search. Integrating any of this into existing offerings might prove costly, and at the same time there may be no pricing pressure with which to offset such costs.
Special Mention: Social Equity
There may be a basic issue of unequal access coming. Yes, of course, this has obviously always been true across a variety of technologies throughout human history! At the same time, at least in the U.S., I think we aspire to somewhat better and we’ve generally had some kind of base layer of universal access for some technologies. Telephony comes to mind as an example. With the rise of the internet from the mid-1990s, I think, (personal judgement), we’ve done a reasonable job of getting ‘online’ into most folks’ hands. Sure, there’s plenty of high value content and insights that are behind paywalls. But the massive corpus of human knowledge available to all is still generally stunning. What I question now is how might GPT technology change this. My series here has been mostly about how GPTs may impact search; not the other capabilities and use cases of AI in general or GPTs in particular. But insofar as they may impact search and its attendant revenue streams, will search maintain its same quality? (I think it likely will. But I also think it’s a fair question.) And will major search engines have an opportunity to offer basic utility type services – GPT or otherwise – to disadvantaged populations? Search is largely free. Or mostly has been. In general, access to asymmetrically valuable information is profitable to the holder. For a variety of such cases, we actually have laws for some aspects of this, such as insider trading laws. Far before such egregious access, we’re all fairly equal in terms of competition regarding open information and we each get to choose how we use it. But now? GPT tools have paywalls on their products. They seemingly have little choice but to choose these business models given their cost structures; at least at present. So access to these tools like generative capabilities and the personal business advantages such access will confer could be a fundamental shift in general information access and usage asymmetry. This article series has been focused primarily on how the general consumer search information retrieval marketplace is managing this latest tectonic shift in technological capability. I’ve not touched hardly at all on the creative / generative powers involved. But clearly, these things favor those who can afford them. Given traditional search will likely all but necessarily have to add some GPT-like capability, it will be interesting to see if it will be enough for those who can’t afford the highest end tools to at least stay in the game; regardless of what game we’re all playing. There is early work on this topic related to online access in general. (See: Digital Divide: Civic Engagement, Information Poverty, and the Internet Worldwide, The Deepening Divide: Inequality in the Information Society, and Technology and Social Inclusion: Rethinking the Digital Divide.
Final Thoughts
Things are changing. Again. I personally agree with those that see the rise of AIs and GPTs as a deep fundamental shift. It’s not a small evolution. It’s on the order of the when we got the web itself, and beyond. Returning to my discussion topic specific to search tool impact… Incumbents should avoid overextending AI, ensuring enhancements complement rather than disrupt core products. GenAI-enhanced search must feel intuitive and natural. While incumbents hold market leverage, success depends on innovation that aligns with user needs. If products ‘outdo themselves’ somehow, they could end up with just a big user experience fail. Traditional search is likely best served by sorting out what it’s best at, communicating that, and hanging on to that part of the market with both standalone and partially integrated tools. Meanwhile, they can compete in new GPT markets using their core search tools for retrieval augmented enhancements.