AI CORNER
The Power of AI to Transform Search
by Barbie E. Keiser
AI is transforming information discovery, as it moves from early adventures with classic AI technologies to generative AI (gen AI) and now agentic AI. It’s not just search, although major shifts are happening there. It’s also browsers, OA repositories, and customized agents that are changing due to rapid developments in AI technologies.
Search engines began the shift by incorporating AI into search results. At first, they experimented with presenting AI-generated responses “above the fold” or in the right panel, sometimes subtly colored to distinguish them from the traditional snippets and blue links. Users can opt for a “traditional” search or select “AI mode” that mimics the homepage display of gen AI tools like ChatGPT.
Browsers assumed a similar trajectory during the past 2 years, overlaying AI on existing browsers or fashioning new ones optimized to deliver answers. Gemini’s floating personal AI assistant summarizes and clarifies content, makes suggestions based on that content, and acts as an agent to execute tedious tasks in the background while users focus on the immediate. (As someone who dislikes pop-up windows, I don’t find the Gemini assistant intrusive; I wonder what Google is doing right that other vendors missed.) An Auto Browse tool that will let AI manage tasks in Chrome is being tested (testingcatalog.com/google-tests-gemini-auto-browse-tool-for-chrome-users/#google_vignette).
Just as search engines mimicked the clean lines of Google’s search page, the AI browser has a minimalist starting page that users have grown accustomed to since their initial encounter with ChatGPT. Along with the familiar browser interface, agentic browsers encourage users to “ask anything” using plain language. These browsers can take actions on the web just like a human, such as clicking and scrolling. They also complete more complex actions, such as summarizing documents, auto-filling forms, booking meetings, or comparing products.
AI-POWERED BROWSERS
Here’s a sampling of familiar names and AI startups to explore as they transition to a new, information-gathering style. New entrants will likely expand features that existing entities must adopt to remain competitive.
Microsoft’s Edge browser incorporates Copilot as a built-in AI companion, allowing users to find websites accessed earlier, summarize information, and perform online tasks. This is possible through Copilot’s access to search histories. Users can enhance their Bing searches by clicking on the Copilot icon at the top of Bing’s landing page.
Mozilla Firefox allows users to employ the AI chatbot of their choice from a selection displayed in its sidebar. Currently, these include Claude, ChatGPT, Gemini, Mistral, and Copilot. Firefox plans a 2026 launch of an intelligence space, AI Window, which will allow users to chat with an AI assistant while browsing. Backlash to Mozilla’s initial announcements of its AI strategy has caused the organization to reconsider whether users will have to opt-in or can opt-out of AI.
Opera first integrated AI into its browser in 2023 with the launch of its Aria browser. Opera’s latest AI browser, Neon, is an agentic workspace consisting of four AI agents: Chat, Do, Make, and ODRA/Deep Research. Opera styles Neon as “an active helper” that works alongside you to amplify the browsing experience.
Perplexity’s Comet Assistant enhances users’ browsing experience by summarizing screen content, finding key data on dense websites, and completing actions, such as scheduling meetings, updating calendars, filling forms, or planning city tours. Access to personal data and preferences is part of Perplexity’s strategy to entice users to abandon their current browser in favor of Comet. Comet can also compare products across open tabs, summarizing the important details in a table. Additionally, Comet can execute research, such as finding a restaurant or flight that meets multiple criteria, noticeably faster than other browsers. Ask any question, and view what Comet is doing on a site in the right panel or let the browser work in the background while you multitask. Researchers seeking reliable citations may prefer Perplexity’s Comet over alternatives. Comet’s free tier lets users access all its browser features plus five searches per month, typically limited to the Perplexity Pro ($20 a month) user, such as background assistants and email integration.
The Atlas browser has ChatGPT built in. The new tab page lets users ask a question or enter a URL. From the ChatGPT sidebar, you can summarize content and analyze data. Browser Memories lets ChatGPT remember context from visited sites. Agent mode interacts with sites, automating tasks, planning events, and booking appointments. The basic features for all ChatGPT options are free, with token limits. Plus or Pro subscribers can add $20 a month to unlock Atlas Agent mode.
Best known for its Arc browser, the Browser Company has transitioned from search and click to contextualizing user intent with its Dia browser and the intelligent workspace it creates. AI features that operate in Dia’s sidebar chatbot include summarization and writing and editing assistance.
DuckDuckGo is not a true AI browser but its Duck.ai chatbot enhances user searches with AI. The Duck.ai chatbot benefits from DuckDuckGo’s reputation for making privacy a priority. Users can select the AI chatbot that best fits the search underway: OpenAI, Meta, Anthropic, or Mistral. Suggested use cases include write code, prepare for a purchase, write an email, and translate text. Norton’s Neo.ai browser claims it “thinks with you,” adapt-ing to your online habits and improving productivity by organizing user workspaces. As you would expect of a Norton Security product, it emphasizes privacy and security. Neo was designed with AI built in, not added on. Its AI Assistance provides suggestions and actions as you browse. It’s Smart Tab Management groups tabs by topic to circumvent tab overload. The browser’s configurable memory feature lets you decide what it will remember of your browsing activities. Neo combines aspects of search, chat, navigation, and actions, which is a new browser concept.
Brave extends its privacy and security protections to its Leo AI app. Leo makes every page interactive. Users can chat with Leo about a topic or ask Leo to summarize a webpage, analyze a document (PDFs, Google Docs or Sheets), or translate text. The free version has fewer models and lower usage limits than the $20 a month Brave Leo AI Premium.
OA REPOSITORIES
Platforms optimized for academic research leverage AI’s strengths, such as summarizing documents, presentations, or websites. Users can query a collection using natural language instead of developing complex search strings of synonyms, related concepts, and keywords connected by Boolean operators, although the systems execute those steps in the background. Analysis can be displayed in formats designed to help the user understand specific aspects of the content, such as tables or visualizations illustrating relationships among documents, authors, or concepts.
AI improves the user experience for repositories whose search algorithms deliver less-than-satisfactory results. For those dissatisfied with Google Scholar’s display of search results, Google Scholar Labs may be the answer. (As you try Google Scholar Labs, watch the left panel as it details what is happening behind the scenes.) “The AI-powered version of Scholar lets users pose complex research questions and receive a curated set of relevant scholarly papers, paired with an AI-generated explanation (a structured summary) of how each paper addresses the query” (ce-strategy.com/the-brief/humain-peer-review).
Scholars access many open repositories to complete the research-gathering stage for future articles. The organizations managing these repositories can use AI to improve the retrieval of relevant material. Here are some examples:
The Allen Institute launched Ai2 Scholar QA (beta) at the start of 2025 to retrieve academic papers from an 8-million-paper repository covering the fields of computer science, medicine, environmental science, and biology. Two months later, the Institute announced the large language model-powered literature search system, Ai2 Paper Finder. At the close of 2025, PaperFinder was rebranded as Asta. One exceptional use case for the tool: “Compare two theories of X.”
Users search arXiv to retrieve papers from its nearly 2.4 million scholarly articles (not peer-reviewed) covering “the fields of physics, mathematics, computer science, quantitative biology, quantitative finance, statistics, electrical engineering and systems science, and economics” (arxiv.org). Two Stanford University students “have developed an extension of arXiv that creates a centralized public square, of sorts, for researchers to discuss pre-prints” (spectrum.ieee.org/alphaxiv). Explore the repository using the now familiar AI interface at alphaXiv (alphaxiv.org) to see the difference in its approach.
CUSTOM AGENTS
AI workhorses can remember what you’ve done and execute routine operations based on experience with past usage. These agents provide domain expertise, extending the functionality of Gemini (Gems), ChatGPT (Custom GPT), and Claude (Agent Skills).
Gems are detailed prompts that instruct Gemini to execute repeatable tasks. Users build custom Gems by uploading files to provide context for future research, such as illustrating the tone and style of writing or expert knowledge about a topic. Greater specificity of instructions in Gems yields more tailored responses for users (blog.google/products-and-platforms/products/gemini/google-gems-tips).
Custom GPTs (chatgpt.com/gpts) are tailored instructions that influence ChatGPT’s responses. Users create custom GPTs by asking the editor to help you build a new GPT that does X. Once you name and describe your GPT, look for the Configure tab to select actions you want “your GPT to take, like browsing the web or creating images” (help.openai.com/en/articles/8554397-creating-a-gpt). This can be as simple as sites you want ignored when conducting research. When you “publish” your custom GPT, you can share it with others—which is particularly helpful when working in teams. Custom GPTs are so easy to create that some users create them, download the file, and upload the instructions to Claude Skills.
Gems and Custom GPTs operate when you call on them; Claude draws on reusable instructions you’ve created when it “feels” the Skill is relevant. User prompts that match a Skill description triggers Claude to load the Skill instructions, metadata, and optional resources (scripts, templates), eliminating “the need to repeatedly provide the same guidance across multiple conversations.” Pre-built Agent Skills provide models for crafting your own (platform. claude.com/docs/en/agents-and-tools/agent-skills/overview).
Users must have a paid version of Gemini, ChatGPT, or Claude to take advantage of these advanced tools. It may be worth paying $20 for a month’s access to each, just so you can see if there are some productivity gains. Do the AI tools deliver more precise results the way you like to see them? Does one do better than another? This seems to be worth a $60 investment to know what you’re missing. |
