I’ve always enjoyed search, both as user and builder. So from a product perspective, I’ve been fascinated by its evolution and the recent fires lit under the traditional tools thanks to the ascendence of AIs. This will be a three part series. First, Search Tools in a GPT world, then business models, and lastly, how traditional search might respond.
So… How might the “traditional” Search industry evolve in the face of AI GPTs? Let’s take a historical tour to consider some customer pain points and values that various tools deal with and how these are morphing. It’s not as simple as GPTs are better search and it might be useful to consider other technology shifts. Did Video Kill the Radio Star? Maybe. But video didn’t kill radio. At least, not completely. Yet. OK, yes, perhaps the shift decimated revenues, but niche use cases survived through both television and even through more recent digital streaming. Even satellite radio was also able to find a place. Will the information retrieval industry experience something similar with what’s been billed as an even more disruptive technology? Or is this truly something radically different if we consider this shift on the level of industrial revolution?
Will the future of Search follow a similar path? Perhaps somewhat, but maybe not quite the death blow some have suggested given there seem to be a lot of niche values for Search. AI driven GPTs, (Generative Pre-trained Transformers), are already changing the search landscape. But their evolution is not as simply obvious as “this is a better search” for at least two related, but separate reasons. First, GPTs can likely excel past traditional search for a wide variety of use cases. But perhaps not all. And second, GPTs can and are used for significantly different use cases than search.
Industry Evolution
First, we need to level set just a bit. What have we had up until recently?
- Search Infrastructure: This is the “picks and shovels.” That is, search platforms, data pipeline tools, etc. This isn’t our primary concern here as we’ll focus more on consumer facing products.
- Major and Minor Search Engines: Basically, the Google Winner-Take-Most option, then Bing, then Brave, DuckDuckGo, followed by various gasping attempts by others.
- Specialty Tools: Low usage, but there are adherents to things like Wolfram Alpha, Giphy image search, audio, others.
- Embedded Retail: Perhaps Amazon should just be its own category, but their primary website is as much about retail search and discovery as it is about purchasing. After Amazon, there’s everyone else.
- Embedded Specialty: Pick a category. Travel, Flights, Hotel, Car Rental, Music, Legal, etc.
What is the Nature of Traditional Search?
Typically, this is keyword or phrase search. Then we have filtering options, sorting, various UI/UX widgets and so on. These providers have become more sophisticated from early days. And yet, the faceted metadata used for product properties, and other digital content assets from imagery to video is still struggling towards best practices. This may be opinion on my part; but seems obvious if you delve into the mess of history from UPC coding to the sloppiness in various other commercial datasets, inconsistent taxonomies across industries, and so on. And these messes range from ecommerce to healthcare records to media to supply chain logistics to science and more. Inconsistent taxonomies create barriers to interoperability, data integration, and seamless user experiences. And yet, we somehow get by. How?
In essence traditional web search for most text information types has two components, and maybe one specialty area:
- Text. The core component, involving keyword matching and relevance ranking.
- Faceted Metadata. Attributes or categories (facets) assigned to content.
- Engineered Features. Manufactured attributes based on rules or use of Machine Learning (ML) tools. (E.g., ratings and reviews can be tool and rule based. Recommended products, category assignment, visual similarity, and more can be done by rule or by ML clustering tools.)
There are others of course; user behavior signals, contextual or temporal data, structures and markup, and more. But the core of basic content search was – and remains – these three major categories.
How Do GPTs Directly Impact Search
GPTs are altering the way people think about and interact with search. For the purposes of this search-retrieval focused discussion, we’re going to leave aside most of aspects of using GPTs as part of creator type workflows. For searching, rather than simply retrieving links based on keywords or phrases, GPTs provide rich, context-aware responses that synthesize information into coherent answers. This changes the dynamics of search in several key ways, at least for certain query types.
- Beyond Keywords: Traditional search depends on keyword matching and link ranking. GPTs interpret query intent, delivering responses often skipping the need for additional filtering or browsing.
- Interactive and Conversational: GPTs allow conversational search, and follow-up questions to refine queries. This allows users to dig deeper without restarting their search.
- Other Data Types: GPTs are often part of larger ecosystems that include image generation, summarization, and more. This can extend to where users get tailored outputs beyond text, such as a travel itinerary, a chart, or a piece of code, without having to visit and evaluate a variety of link targets.
- Search as a Platform for Creativity and Exploration: Rather than confined to existing content, GPTs open search as a creative tool. For search use cases where users were just going through links for research while crafting something new, now a user can go right to the creation process.
- Democratizing Expertise: Traditional search often prioritizes authoritative sources through algorithms, but GPTs can synthesize insights from niche or less prominent sources, providing diverse perspectives. This may reduce reliance on dominant platforms while empowering users with broader viewpoints. Though it could arguably also increase risk of outright errors.
In essence, GPTs transform search from a tool for retrieving information to an interactive process of generating and refining knowledge. While this brings new possibilities, of course it bears mentioning that the tools also raise critical questions about accuracy, monetization, and user trust.
What About Brand Lock In and Emotion?
Let’s leave aside features, functions, benefits for a moment. Let’s consider things from a brand and loyalty perspective. Generally, brand lock in is good. Powerful brands are intangible assets potentially worth billions of dollars. Of course, this can be a negative. If General Motors started selling burgers, (even if they were awesome), this might not fly well. Similar to how smell affects taste, you might actually get a ‘tinny’ taste biting into one just from the strength of your mental models and associations. (Though I’d bet Tesla could somehow pull this off somehow. Can you imagine how large a CyberTruckBurger would be?)
Anyway, back to Google… just what they are is likely deeply locked in users’ psyches. Even as they start to integrate GPT-like tools into their product, users may not adopt them. And if Google tries to change too much to match the new upstarts? They may rattle users’ cognitive models and at the same time deprecate some of their other valuable features. Double whammy. They could reinforce an idea we’ve all likely experienced, “You know, that place just isn’t quite the same as it used to be.”
There also may be some anti-monopoly sentiment in consumers. Few of us trust “too big.” I can’t prove this, but I think in the early days of search, people turned away from Microsoft simply because they thought, “you have enough,” and it was nice to have a bright, shiny new thing in Google. And now Google might be feeling a little dusty. Again, this is opinion, but I believe that sometimes what large brands perceive as customer loyalty is really more learned helplessness due to lack of options. Consumers aren’t loyal; just trapped. The larger the brand and the stronger the lock in, perhaps the more arrogant the marketing and product leadership. And riper for disruption. Why? It’s not just because the new thing can catch them with their pants down and bureaucracy up. It’s because even if incumbents can change features and functions, undoing long baked in mental associations is much more difficult.
Some have managed pivots, generally somewhat adjacent. Netflix may have been killing Blockbuster with DVD by mail prior to online, but they still had to jump to streaming. And they managed that. Apple jumped from just computers to launch of the iPod, iPhone, and iPad. In fact, we still call digital radio shows podcasts even though the iPod is no longer sold! (Their brand meme lives beyond the product sunset. That is lock in.) There’s other examples. But successful pivots are hard. Packaging is critical. Trust may be transitive, but people can and will only jump so far.
So What Will Become of Traditional Search?
While traditional search tools will incorporate both Machine Learning and GPT tools, a lot of their value propositions will likely become more niche, focusing on what they do best. Traditional search isn’t likely to wholly disappear into the dustbin or history. Its role will evolve. As GPTs reshape how users interact with information, search engines will adapt by leaning into their unique strengths, refining niche value propositions, and integrating AI technologies without abandoning their core functions.
Where can traditional search engines perhaps maintain supremacy in fit-for-purpose areas?
- Faster and Broader Indexing
Traditional search excels at rapidly indexing and retrieving vast amounts of data. While GPTs generate content dynamically, search engines will provide real-time access to the latest information. GPTs will try to to this, and offer hybrid solutions. But struggle due to reliance on pre-trained data. I’d expect search engines to double down on their advantage in speed and breadth, ensuring fresh content remains a differentiator. - Fact-Checking and Verification
GPTs can hallucinate or confidently present incorrect information. Traditional search can position itself as a verification tool, offering authoritative sources to cross-reference AI content. By becoming a trusted validation layer, search engines can emphasize source transparency and accuracy. - Specialized Search Domains
Expect growth in vertical search catering to specific industries (e.g., legal, academic, financial). Traditional search can offer deep, curated datasets that GPTs may not easily access or synthesize accurately. This specialization will drive niche markets where precision and source credibility matter most. - Hybrid Models
Search will increasingly integrate GPT-like features, blending dynamic answers with traditional results. This hybrid approach maintains the benefits of exploratory search while introducing AI-generated summaries, or other snippets on result pages. The balance between curated links and AI results will create a more comprehensive user experience. (I’m planning to write more about this an upcoming article about how traditional search might respond to the rise of GPTs.) - Focus on Visual, Local, and Retail Search (Among other niches.)
Areas like image, video, and local search remain difficult for GPTs to dominate without extensive real-time data. (Though they continue to grow as creation tools.) Traditional search will likely pivot towards enhancing visual recognition, geolocation services, and retail discovery, where structured data and user-generated content matter. And while ML techniques will likely be used for some data types, GPTs are not necessarily required. - User Control and Filtering
Traditional search offers users control through filters, advanced operators, and sorting tools, allowing for refinement. This contrasts GPTs’ black-box nature, where results are via the model’s internal logic. Natural language may be a useful interface. But it can have interaction challenges of its own. Emphasizing user agency and filtering might differentiate search engines as tools for research, deep dives, and data mining. - Ad-Driven Economy
While GPTs challenge ad-based models, traditional search will continue to monetize through targeted ads, sponsored results, and affiliate links. The ability to direct users to transactional opportunities from products to services ensures search remains a dominant force in ad revenue ecosystems. (I’ll also get into this a bit more deeply in a business model specific article.)
Final Thoughts
Traditional search isn’t going away just yet. It will need to pivot at least somewhat. And a lot of category revenue may shift elsewhere unless the whole pie grows significantly. By focusing on real-time data, accuracy, niche specialization, and hybrid AI integration, search engines will carve out enduring value alongside GPTs, coexisting rather than being completely replaced. But there will be some suffering along the way.
So What’s Next?
We should look at Business Model differences first. Then consider how Traditional Search can respond. Those will be a couple of upcoming articles..
If you liked this article, consider these earlier thoughts on: How Badly Will the GPTs Kick Google’s Teeth In.