How LLMs Will Change Web Browsing and Web Publishing
How LLMs Will Change Web Browsing and Web Publishing
Opera's new announcement shows LLM features will be common in web browsing
Opera announced that it would soon add generative AI tools to its web browser. The company only had a 2.4% browser market share in January 2023, according to Statcounter, which is just a couple of percentage points behind Microsoft Edge’s 4.5%.
Opera almost certainly sees these tools as a potential opportunity to steal market share away from Google Chrome, but just as importantly, the technology poses a threat. Adding large language model (LLM) and generative AI features may soon be necessary to ensure existing users don’t bolt for another browser.
There is little doubt the company has been working on this for some time. In addition, the rapid rise in Edge browser downloads the day after Microsoft’s event last week was surely noticed by Opera executives. That made it all the more important to announce the new features even though they are not yet available. No company executive wants to miss this moment. Song Lin, Co-CEO of Opera commented:
“Following the mass interest in generative AI tools, we believe it’s now time for browsers to step up and become the gateway to an AI-powered web”.
Summarization as a Gateway
Summarization and conversational interactions with web content are not the only features Opera plans to bring to the browser. Like Microsoft, the company indicated it will also add generative writing features with summarization as the first to roll out.
Per Wetterdal, Head of Strategic Partnerships and AI ecosystem, also said he is excited for the “roll-out of developer programs for solutions such as Google Bard.” This suggests Opera is starting with OpenAI’s GPT-3 but will be open to using a variety of solutions.
However, the first feature will be GPT-3 powered web page summarization. The company characterizes this as a necessary countermeasure to the expected rise in content generated by LLMs. “Users of the Web are constantly flooded by content in amounts that are impossible for us to process. With generative AI helping us create written materials faster, there will only be more and more content.” How perfect. The problems the LLMs create can be solved by LLMs!
How Will LLMs Change Web Browsing
Lin’s Opera colleague, Krystian Kolondra, EVP PC Browsers and Gaming, added:
“It is Opera’s belief that with AI solutions springing up both for text, image, and audio generation and in countless other forms, we are at the brink of a new era of creativity on the Web. That’s why we’re reshaping the browser to allow our users to tap into all these resources.”
There has been a lot of talk about how LLMs will change writing and search. Far less has been said about how it will change web browsing. Summarization and conversational interactions specific to a web page will no doubt be popular. The example cited by Opera is a news article. That is one scenario where getting a condensed summary could save the user time and improve their web browsing experience.
Another scenario is more of a search feature for the website. At a basic level, it could be simply finding something on a particular page. It is not clear that LLMs will be better than standard retrieval models on these smallish data sets with constrained domains. However, the ability to have a conversational interaction about the content of a website would be very helpful.
Chatbots have attempted to fill this gap on many websites but generally fall short of user expectations. Very often, the question a user has does not align precisely with the top-level messaging that marketers feel compelled to put in front of every website visitor.
Navigating a website can be tedious as layouts are not common, and many companies don’t have a strong grasp of the long-tail questions of prospective customers. They also tend to atomize products in ways that sometimes obscure how a user can benefit from a company’s products. A conversational search option could help users get answers faster and provide website publishers with more information about what users really care about.
It is unclear whether LLMs will make up for the inadequacies of today’s chatbots. Surfacing content in constrained domains differs from open domains, where LLMs have proven so deft. However, the step-change performance of solutions such as ChatGPT has created optimism and confidence that performance can become radically better.
How LLMs will Change Web Publishing
It also seems clear that these changes will also impact web publishing. Today, the dominant drivers of web design strategy are brand, customer persona, buyer journey, and Google search engine optimization (SEO). SEO becomes the key driver for many companies, given that the other factors only come into play after you have a website visitor.
Very soon, we may see every website need to consider how and when it gets surfaced as a source in a Bing or Bard chat answer. If conversational search achieves mass adoption, conversational search engine optimization (cSEO) will be critical for driving organic discovery.
Keep in mind the ten blue links model will still exist, but if consumers begin using conversational search more, the citations and quick links that drive traffic are going to be more limited. There is not that much real estate to show buttons or inline citation links that will drive website traffic.
Another consideration is how citations appear in answers. Today you get a blue link to your query. Conversational search provides an answer that may have several different elements. Each element can have one or more citations. A key benefit is that the user then knows what information and context each link addresses. This should help drive much more value for having very specific as opposed to general information on a web page. You may not receive the best overall webpage but instead the best-predicted webpage for a specific idea within a response.
Finally, we come back to summarization. Every marketing team and SEO specialist will need to test websites with the browsers to see what elements they extract for the browser summarization feature. This is likely to lead to a new kind of website copywriting, and it may be hard to optimize for each browser’s implementation as they will all tune their models differently. It will also require regular monitoring, much like companies do today with keyword analysis.
Given this new model of web browsing, you can expect the SEO software companies, such as Moz, ahrefs, and SEORush, to soon roll out generative AI features to track results in conversational search and browser summarization. Google Analytics may also introduce a Bard-specific tool for this. Of course, the SEO software players are also likely to introduce AI writing assistants that optimize keyword content production. That is an obvious feature for them to add, and everyone from OpenAI and AI21 to Jasper AI will be happy to accommodate them.
The Web Browser Wars
Opera and Edge have made their announcements about new browser features powered by LLMs. Chrome, Safari, and Firefox are likely to follow suit. This may lead to a new round of browser wars, but if every browser adopts the same features we are more likely than not to see the status quo prevail. Microsoft is the one company that seems well-positioned to gain market share because of its other services.
Consumer habits typically don’t change very quickly, so the initial adoption rate for conversational search and webpage summarization through browsers may not happen overnight. However, the value is compelling, and the conversational interfaces require little or no training to adopt. These features seem likely to be popular and establish new consumer habits around web browsing.
A knock-on effect will be a disruption in web publishing. That might lead to even bigger changes because it will impact so many web publishers and further transform the web browsing experience.
