© IE Online Media Services Pvt Ltd
Tags:
Journalism of Courage
When OpenAI CEO Sam Altman announced ‘SearchGPT’ last week, he dismissed Google as an ‘old-school search’. This indicates just how bullish Altman is about his new AI-powered search even though it is currently just a prototype. While SearchGPT is far from a public release and is currently being tested with a small group of users, the extent of the impact it could have on Google’s $175 billion-a-year search business is yet to be seen. However, OpenAI’s approach to search is similar to many that have tried to emulate and spectacularly failed at a time when search itself is changing.
Although OpenAI did not offer a hands-on demo of SearchGPT, it previewed the feature via a prerecorded video. The interface and functionality resemble traditional search engines like Google, with a large text box that prompts users to enter their queries. However, instead of presenting plain links like Google, SearchGPT uses GenAI to compile links and answers with clear and relevant sources. In one example provided by OpenAI, the search engine summarises results on music festivals, presenting short descriptions of the events and including attribution links. In another example, it explains when to plant tomatoes before breaking down different varieties of the plant. After the results appear, users can ask follow-up questions or click the sidebar to open other relevant links.
In a way, SearchGPT combines elements of traditional search with GenAI features, similar to Google’s AI Overview, which faced public criticism after users spotted inaccurate results. The similarities don’t end there; SearchGPT also presents search results using real-time access to information on the internet, modeled on Google search. Additionally, the search AI tool will utilise information from publishers who have signed deals giving OpenAI. Yet again, the approach is modelled on how search works on Google which partners with news publications to get precise and accurate results. Eventually, OpenAI aims to integrate SearchGPT directly into the ChatGPT AI chatbot.
There is nothing wrong with emulating Google’s business model and creating a search product that works better than Google. “We think there is room to make search much better than it is today,” OpenAI CEO Sam Altman wrote in a post on X. Competition is always welcome, as Google has long enjoyed clear dominance in the search business and continues to do so.
It’s not that no one has tried in the past to challenge Google’s search; attempts have been made, but competing with Google has proven difficult. The list of companies that have failed to post a serious challenge includes tech majors like Yahoo and Microsoft, as well as startups such as Neeva, which folded in 2023. Once poised to challenge Google’s supremacy, Neeva’s co-founders, Sridhar Ramaswamy (a former head of Google’s ad business) and Vivek Raghunathan, attempted to create a superior ad-free, subscription-based search engine but could not convince users to switch from Google. “We’ve discovered that it is one thing to build a search engine and an entirely different thing to convince regular users of the need to switch to a better choice,” the co-founders wrote in a blog post announcing Neeva’s shutdown.
Similarly, Microsoft’s latest attempt, even after giving its search engine Bing an AI makeover, has increased its share but still not made a major dent in Google’s search dominance. Statistics from market intelligence firm Statcounter show that Bing’s global search engine share stood at 3.72 per cent in June, while Google was miles ahead with 91.06 per cent. Smaller competitors like DuckDuckGo and Brave, which both market their search engines as privacy-centric, remain in business, but they too aren’t large enough. This leaves OpenAI and Perplexity AI, another well-funded startup that pitches itself as an AI ‘answer’ engine, as formidable forces aspiring to capture a slice of Google’s lucrative search business.
However, OpenAI’s search ambitions face the same bottlenecks as every other player who came before it and fizzled out without coming close to Google in perception and market share. The hard truth is that a search engine needs an index of the web, and Google has a comprehensive map of the internet built over two decades making it the power it is. Simply put, before any search engine can even claim to be a competitor to Google, it has to crawl the web. Because the web is vast, this is a significant challenge.
Most websites, especially large ones, are finicky about who can crawl their pages. Google and Bing get more access because they have more users, while smaller search engines often need to request permission. Websites that provide greater and more frequent access to Google’s web crawlers—essentially computers that seek out information on the World Wide Web—are more likely to get indexed, which translates to higher visibility and ranking.
This mutual relationship is why websites and search engines depend on each other. Websites rely on search engines for traffic, while search engines need access to crawl sites to provide relevant results for users. This is how Google search works, setting the standard for traditional web search. Even if a search engine excels at providing web results, it will struggle to match Google’s non-search business, which includes displaying ads and a spectrum of services like Google Analytics.
Another way Google controls the web is through metrics; it knows which websites are popular and how big they are. For a small search engine, replicating this is challenging. A startup like Perplexity AI may boast about how much better it is than Google, but its market share is so small that it is virtually unmeasurable.
While OpenAI aims to compete with Google in its core search business, a growing trend is that search is no longer solely about Google. Google’s data suggests that 40 per cent of young people prefer using TikTok and Instagram for search instead of Google. Even though Google still dominates the web and controls the indexing of websites and blogs—thereby providing the fastest search results—its logic applies to traditional search methods, not to how search happens on TikTok or Instagram. The traditional search typically involves typing text, but on a dating app like Tinder, swiping right or left on a profile may be considered a form of searching. This is not how Google defines a search. Similarly, more and more people use TikTok to look up information, such as news.
Unlike Google, which directs users as quickly as possible to sites where they can verify the information they are looking for, TikTok does not aim to redirect users to other sources of information. Instead, the app wants users to spend as much time as possible to increase engagement. It recommends videos that have a lot of reactions or are trending, which may explain why so many young users are flocking to TikTok or Instagram for news instead of searching for publications on Google.
With SearchGPT, OpenAI is adopting a similar approach to Google, such as leaning heavily on its relationships with publishers (Perplexity has struggled in this regard and is accused of copyright infringement for scraping online articles without permission to feed its AI). However, OpenAI will have to explain how it differs from traditional search engines like Google. The AI search aspect that OpenAI is banking on as a differentiation from typical Google searches has either been inaccurate or not as helpful as claimed. The early promise that AI search would revolutionise how we navigate the internet has been exaggerated, as demonstrated by the failure of Google AI Overviews. Nevertheless, Google remains dominant on the web, even as search and user behaviour evolve.
