The Future of Search Hinges on How Much We Trust AI
What are Arc Search and Perplexity, and what they mean for the future of online search & discovery
Ever since Microsoft unveiled a new Bing search interface powered by ChatGPT over a year ago, there has been an arms race among tech giants and startups to reimagine the search experience by integrating advanced generative AI capabilities.
Early on, thanks to its deep partnership with OpenAI, Microsoft’s Bing emerged as a frontrunner to challenge Google’s dominance in search, forcing Google to play catch-up, fast-track the rollout of its Search Generative Experience (SGE), and most recently, rebranding Bard to Gemini to consolidate its AI products. Yet, a few months into the race, it’d become clear that ChatGPT-powered Bing was not chipping away Google’s market share in search.
Last week, however, a buzzy newcomer joined the race. The Arc Search app, created by The Browser Company, which is also behind the alternative web browser, Arc, introduced a new feature called “Browse for Me,” which takes a novel approach towards presenting search results: Instead of the pages of blue links you’d get with Google, or the type of generalized, source-less summary returned by the latest crop of AI chatbots, Arc Search’s standout feature will streamline the search experience by automatically gathering and summarizing relevant information from at least six different web pages, and then presenting the results in the form of a custom-built web page with key takeaways listed at the top. Here is an example:
This AI-powered search service, available now via Arc Search’s iOS app, marks a significant move towards redefining the search engine experience. Positioned as a more intuitive solution for AI-driven web searches compared to existing tools, it offers a unique user experience by synthesizing web content into concise, easily digestible summaries of the information they seek. The underlying AI models powering this feature come from a mix of OpenAI and other AI companies, according to The Verge. However, considering the timing of the launch, it seems safe to assume that Arc’s recent integration with Perplexity, an AI-driven search engine, also played a role.
For context, Perplexity is one of the leading AI “answer engines” that popped up last year. Instead of serving up traditional search results, Perplexity uses AI to generate summaries of online content to answer inquiries. Its key features include the ability to source from a specific domain, such as academic papers, YouTube videos, or Reddit posts, as well as a “Copilot” feature (no relation to the Microsoft Copilot) designed to help users narrow down a query by asking clarifying questions. According to the New York Times, Perplexity runs on OpenAI’s GPT-3.5 model along with its own AI model built on Meta’s open-source Llama 2 model, and it is free-to-use. However, users can also upgrade to a Pro tier for $20 a month and gain access to more powerful AI models, including GPT-4 and Anthropic’s Claude, along with some bonus features, such as the ability to upload their own files.
By integrating Perplexity into the Arc Browser and the Arc Search app as one of the many options for default search engines, The Browser Company leverages cutting-edge AI technologies to redefine how users interact with the internet, offering a more efficient and intuitive search experience. And one could easily see the inspiration behind the “Browse for Me” feature, which, to be clear, is powered by Arc’s own AI models.
All in all, it is clear that the competitive landscape for search is quickly evolving, with exciting new approaches popping up to push the envelope on our search and browsing experiences, which are trending towards more intuitive, AI-driven search experiences, challenging traditional search paradigms and, by extension, how the digital media ecosystem works.
A Disruption in Search of a Business Model
One of the immediate reactions to Arc Search’s latest feature is a pessimistic view of the future of the web. The digital media economy has always relied on search as a key lead generation. As Pranav Dixit succinctly puts it in a piece for Engadget:
For decades, websites have served ads and pushed people visiting them towards paying for subscriptions. Monetizing traffic is one of the primary ways most creators on the web continue to make a living. Reducing the need for people to visit actual websites deprives those creators of compensation for their work, and disincentivizes them from publishing anything at all.
If emerging AI search experiences like Arc Search’s “Browse for Me” feature and Perplexity’s AI search were to become a popular way for people to access online information, where A.I. does the browsing for you and serves up all the information you need right on an easily digestible answer page, one could imagine that most people would not bother clicking through to the source sites anymore. With reduced traffic comes diminishing revenues for digital publishers, and that spells a death spiral for the already floundering digital publishing industry. Even CEO of The Browser Company, Josh Miller, acknowledges the disruptive potential of generative AI on the web’s economic model, but admits uncertainty about how to compensate original content creators.
Within this context, it’s easy to see why many digital publishers are panicking about AI search right now. Some publishers are fighting back, including the New York Times, which sued OpenAI and Microsoft for copyright infringement last year. Still, some are inking deals with AI companies to explore potential content partnerships to feed their content into the various AI models for a hefty licensing fee. For example, in July, Associated Press teamed up with OpenAI to explore generative AI use in news. Meanwhile, Apple is reportedly talking with some big news publishers like Condé Nast, NBC News, and IAC about licensing their archives and using that information to help train its generative AI model.
To be fair, even before the arrival of generative AI, search engines had increasingly returned responses tailored to the user’s query intent and altered the impact of SEO tactics on gaining clickthrough. Even in 2020, two-thirds of Google searches on mobile ended without a click. If anything, user behavior was already shifting from clicking through to in-platform experiences. Through this lens, this new wave of AI-powered search engines is simply an accelerant of an existing trend.
Still, someone has to do the research, write the breaking news, and create new content, otherwise the AI-powered search engine would have nothing new to feed on. Without that, all we have left would be nothing but AI-generated content, and soon enough AI would start eating its own tail and regurgitate outdated or hallucinated content, to disastrous results. That does not sound like the type of AI search experience that any user would want.
Therefore, for this type of AI search to sustain itself and to grow, a new paradigm of compensating journalists and other content creators must be created to ensure that new content is being created with or without the publishers acting like the middle-man. At the end of the day, AI-generated answers are only as accurate as the sources that the AI models are trained on.
One could theoretically imagine a future where AI search engines like Arc or Perplexity charge users a subscription fee to use their services, and use part of the revenues generated from subscriptions to compensate people who created the latest content that went into training their latest AI models. If that is indeed the future of search, then the place for brand advertisers would become a lot more precarious. With AI search engines providing summarized content, opportunities for direct engagement on publisher sites could diminish. In response, brands will likely need to explore innovative strategies, such as integrating with AI platforms or developing more compelling, AI-friendly content that these engines prioritize.
Luckily for brand advertisers, that hypothetical scenario is nowhere near coming true yet. For one, even though Perplexity has 10 million monthly active users, an impressive number for a young start-up, it is nothing compared with Google search’s billions of users. Perplexity also lacks a lucrative business model. Right now, it has no ads and fewer than 100,000 people paying for the premium version, far from enough to generate enough revenue to fund the future of online journalism.
Circling back to Arc Search’s “Browse for Me” feature for a second: the brilliance of generating a new web page with glanceable information for each search inquiry is that those result pages may serve as the perfect vehicle for delivering a brand message or targeted offer, perhaps even a banner ad or two. For example, when I asked Arc to “Search for Me “about what products to buy to help deal with insomnia, Arc Search quickly produced a page full of popular sleep aid products.
If your brand happens to sell any sleep-adjacent product, then that would be the kind of search result where your brand can and should show up, preferably with a shoppable link. That said, given that this type of in-platform, frictionless search experiences are not designed to encourage clickthroughs, direct-response ads may not even work that well for them.
Make no mistake, for this new search experience to go mainstream, it would require a seismic shift in how the entire digital media industry operates and monetizes its content. Until that happens, AI search will remain a neat product in search of a viable business model.
The AI Trust Gap
Another obvious but crucial adoption hurdle for AI search lies in how much users can trust the search results and summaries that the AI-driven search engines serve up to them. At the moment, most of the AI “answer machines” are not very good at citing their sources. Newer AI search experiences like Perplexity or Arc Search are more careful with their source citations, but still, that may not be enough to quell the concerns of some users. And even with the sources carefully cited, users may wonder about the relative trust-worthiness or, in some cases, the potential political bias, of each source. These nuances inevitably create a trust gap between AI and users, making it difficult for users to evaluate the accuracy and credibility of the information they’re being presented with.
Anecdotally, testing the Arc Search app myself, it is amusing to see the built-in Perplexity search engine return a nonsensical results for a simple search inquiry “who is the first Asian Oscar winner for acting?” by insisting that Michelle Yeoh has not won an Oscar yet, whereas Arc’s own models powering the “Search for Me” feature returned the right answer, but still prominently featuring Yeoh’s Oscar-winning photos atop like a bad misdirection.
Clearly, AI search still has a long way to go to earn the consumer trust.
That said, Gen Z is most likely to be the generation to close that trust gap and embrace AI search. For one, as digital natives, Gen Z is naturally distrustful of online information and better equipped to discern misinformation than older generations. Axios reports that 69% of Gen Z students said it is somewhat or very easy for them to distinguish real news from misinformation, citing a survey from polling firm College Reaction. In addition, studies have found the youngest American adults are far less likely to share misinformation online than are older Americans, despite, or perhaps because of, them increasingly using social media as a primary tool for searching and discovering new information, products, and services.
On the other hand, Gen Z is also driving the early adoption of generative AI, especially in the context of education and career prospects. Unlike some workers who may feel apprehensive about the impact of generative AI on their jobs, many college students and recent graduates view these technologies as tools to gain a competitive edge in their careers. This group, often referred to as “AI natives,” is actively seeking out AI courses and planning to integrate these skills into their future job roles.
That said, bridging the AI trust gap is essential for the future of search. With OpenAI recently rolling out a new Memory feature for ChatGPT to deliver more personalized answers based on the information shared in prior conversations, it also opens the door for potential data privacy issues that could further widen the trust gap. Closing the AI trust gap requires concerted efforts from AI developers, policymakers, and the broader digital media industry.
For brands, this means navigating a complex landscape where transparency and authenticity become more than just value propositions; they’ve become critical components of brand strategy. In an era where AI-driven search engines and chatbots are in the ascendant, ensuring that advertising content is not only visible but also trusted by consumers is paramount. Brand marketers need to adapt by leveraging AI in ways that enhance the credibility and relevance of their content. This involves closely monitoring AI developments to understand how these technologies curate and present information, ensuring their messages align with the platforms’ standards for accuracy and transparency.
