The transformative power of genAI in web searches

The technological revolution brought about by ChatGPT – the large language model (LLM) that might not be better at writing than humans, but is certainly faster – has resulted in some startling related technology that has been looking for use cases ever since.

To understand and answer questions, ChatGPT must have NLP processing, understanding and generation capabilities that extend well beyond the chatbot use case and can be leveraged to create different types of original content as well. Depending on the nature of tokens, this can be – among other types of output – texts, music, videos or code.

GenAI has been trained on a relatively large body of data and is therefore able to access a huge knowledge base of information. That means another natural use-case for generative AI is as a search engine that can answer natural questions in a conversational manner – a functionality that positions it as a potential competitor to established web browsers.

The impressive data retrieval abilities and summarisation performance of ChatGPT and generative AI in general is what inspires journalists and vendors alike to talk in the same breath about how vector-based search outperforms old-school password-based search and spells the end of Google’s monopoly in the web search space.

By doing so, however, they merge two distinct watershed moments of search technology development: one that enables search engines to deal with synonyms and misspellings while still returning an accurate response; and the other that allows us to interact with these engines in a natural human language.

Although Google is undoubtedly facing unprecedented competition in the search engine space thanks to the rise of genAI-powered search, vector search – a semantically aware question-answering system – is not the secret sauce that turns businesses such as OpenAI, Perplexity AI or You.com into challengers.

Vector search – the unsung hero

While search engines with incorporated GenAI capabilities are often praised for their ability to understand context and interpret intent behind a query, these features are not unique to them. Generative search only adds a new layer of functionality to the original dichotomy of keyword and vector search.

More straightforward keyword-based search provides answers by seeking documents that have the highest number of keyword matches with the query. In this case, the engine looks for exact matches and won’t bring up answers that don’t contain the keyword.

Although keyword-based engines are getting better at understanding misspellings and incomplete expressions, their so-called fuzzy matching is far from being their forte.

They can, however, deal with precise queries better than semantic searches, which can be key in an e-commerce context, where shoppers may want to search model numbers or specific brands instead of product categories. Keyword-based engines are unbeatable when the user knows what they want.

The feature of vector or semantic question-answering systems that revolutionised search is their capability to index unstructured data, ranging from text to audio-files, videos, to social media posts, webpages or even IoT sensor data.

Vector search also plays a central role in genAI model training, as well as by enabling these models to discover and retrieve data with impressive efficiency.

But anyone that has been using Google search – roughly 90 per cent of all web browser users – can attest to its excellence at coming up with relevant answers to queries no matter how misspelled, fuzzy or ambiguous they are.

This is because Google developed the fundamental concept of vector search into a powerful tool and has been honing it through various updates ever since.

How integral Googling is to the human experience today is also highlighted by the argument that advocates of bionic implants use to fend off criticism by calling the Google browser a “mental prosthetic”. In the old days, you quizzed people around you about the snippet of information that escaped your memory; today, you google, ChatGPT, perplexity or you.com it. But does it matter which one you use?

Generative search

While the generative subset of AI has been stealing the show from other, more established types of machine learning algorithms, it, in fact, leverages another strain of the transformer architecture that Google uses since the BERT update of its search engine in 2018.

Search engines can improve by the day, which applies both to vector and generative search, and there are plenty of articles and test sites that compare Google’s and ChatGPT’s ability to read user intent and its nuances, as well as how accurate and reliable the search results are.

However, for ChatGPT, usage as a search engine is only one of the use-cases – and a secondary one at that. What makes generative AI pivotal, and also enables it to take the Google Search universe on (Google maps, Google reviews, YouTube and all) is the new way in which it accesses information.

By not simply providing citations that the user can extract a query answer from, but generating a human-like response that synthetises an answer from the most relevant information snippets found in the model’s training data, generative AI sets the bar higher.

Consequently, there are two levels where it can fail – by not identifying the most relevant and accurate information sources, and by not generating the right answer from the top search output.

Even if it retrieves the right data, the generated content that it delivers as a reply to the query may contain inaccuracies or prove a fabrication – confident and authoritative although it may sound.

Conflicting information that ChatGPT encounters in the sources that it draws query answers from, for example, can result in lack of coherence and contradictions.

Authenticity is another issue: earlier LLMs couldn’t cite where they drew their answers from. New generative search models, however, are increasingly addressing these pain points.

Perplexity, a leading answer engine, for example, while also making the most of content generation by offering an automated news feed functionality, also shows the top sources where elements of its generated content have been extracted from.

E-commerce site search and enterprise knowledge platforms

Although Google will probably remain the go-to information resource while the additional value of conversational, intuitive interfaces isn’t enough to offset concerns about their authenticity, generative search is expected to disrupt the status quo in the online customer and the employee experience space, where emphasis is on the experience. These areas also have the advantage that they can ground LLMs in use-case specific information – in a website or a company’s enterprise database, thus drastically reducing the risk of  hallucination.

There are convincing use-cases in both realms. Online shoppers often find that googling a product and the retailer in a combined search will take them to the product much more efficiently than a product search leveraging the e-commerce site’s own tool.

Searches on enterprise data also have a reputation for being time-consuming. Although there is some variation between the findings of different studies, general consensus is that knowledge workers spend too much time retrieving information from enterprise data bases.

This problem is further aggravated by data silos, and the fact that employees on average need to access four or more software systems to find the information they need to complete their tasks. Generative search’s ability to deliver instant answers in a style mimicking human conversation has the potential to link each online shopper with a competent virtual shop assistant who will spare them endless virtual ramblings through a maze of tabs and drop-down menus – which often lead to cart abandonment.

Meanwhile, in the office context, generative search tools can provide each employee with a savvy work-buddy whose knowledge spans across all functions and departments.

The second part of the series will explore the potential that generative search presents in reducing cart abandonment and improving productivity at the workplace, as well as the challenges of implementation.