The digital landscape is undergoing a profound transformation with the advent of conversational AI search engines, tools powered by large language models (LLMs) designed to retrieve, summarize, and present information from the internet in a dialogue-like format. This paradigm shift, highlighted by a February 2024 report in Fast Company, represents a significant departure from the traditional keyword-based search model that has dominated the internet for decades. As academic and scientific research actively explores this burgeoning field, a wave of experimentation is underway, fueled by fervent enthusiasm for what promises to be a fundamentally new interaction model for information discovery.
The Dawn of a New Search Paradigm
For over two decades, search has been synonymous with Google: a query box, a list of blue links, and the user’s implicit task of sifting through results to find an answer. This established mental model, deeply ingrained in the habits of billions of internet users, is now being challenged by generative AI. The emergence of powerful LLMs, exemplified by ChatGPT’s public debut, has ushered in an era where AI can not only understand complex queries but also synthesize coherent, human-like responses. This technological leap has spurred the development of new search applications like Perplexity AI and Andi, which aim to move beyond mere link aggregation to direct answer provision.
The shift is not merely an incremental improvement but a fundamental re-evaluation of the user’s journey in seeking information. While the core user need – finding answers – remains constant, the methods for achieving this are rapidly evolving. These conversational AI engines embody a "technology push," where a potent new capability, generative AI, is driving market innovation and shaping digital experiences. However, the true extent to which this evolution will universally benefit users, and the potential strategic risks of prioritizing technology over human-centered design, remain critical areas of inquiry.
Redefining User Interaction: A Shift in Mental Models

At first glance, the home pages of conversational AI search engines like Perplexity AI retain familiar elements: a prominent input field for queries, a central display area for results, and supplementary widgets. Yet, the underlying mental model is radically different. Unlike the traditional search engine results page (SERP), which presents a curated list of external links, Perplexity aims to directly answer a user’s question by collating and summarizing information from multiple sources. The user experience is less about navigating a directory and more about engaging in a conversation, akin to interacting with a chatbot. This conversational interface has become front-of-mind for many users due to the widespread adoption of tools like ChatGPT, reinforcing a new expectation for AI interactions.
Andi, another innovative search engine, pushes this conversational mental model further in its layout and interaction design. While its visual interface leans heavily into a chat-like format, its information architecture, as will be discussed, maintains some conceptual ties to classic search, offering an interesting hybrid approach. The core distinction lies in the AI’s role: instead of serving as a sophisticated indexer, it acts as an intelligent intermediary, interpreting the query, synthesizing information, and presenting a cohesive answer. This direct answer format dramatically simplifies the user’s task, removing the need to evaluate individual search results, make educated guesses about content relevance, or manually scan through multiple web pages.
Enhanced Usability: Benefits for the Modern Searcher
Conversational AI search engines offer clear advantages in usability and interaction design. By delivering direct, synthesized answers, they streamline the information retrieval process. For instance, a query in Perplexity AI yields a text-based response that mirrors the format of an article or web page, rather than a list of hyperlinks. This approach directly addresses a crucial usability heuristic: matching the system to the real world. Human communication often follows a question-and-answer pattern, and these AI engines replicate this intuitive interaction, making the search experience feel more natural and efficient.
In the context of Jakob Nielsen’s widely recognized usability heuristics for user interface design, AI search engines demonstrate several improvements. They enhance "user control and freedom" by reducing the cognitive load associated with navigating search results. They promote "recognition rather than recall" by presenting immediate answers instead of requiring users to remember which links might be most relevant. Furthermore, they offer "flexibility and efficiency of use" by allowing both novice and expert users to quickly obtain information. The system’s ability to "help users recognize, diagnose, and recover from errors" is also implicitly improved, as the AI’s direct summary aims to anticipate and fulfill the user’s intent more accurately from the outset.
The speed and convenience of receiving a concise, pre-digested answer represent a significant efficiency gain. Users no longer need to spend time evaluating the credibility of multiple sources, clicking through various links, or meticulously extracting information. This directness aligns with the fast-paced nature of modern information consumption, promising a more rewarding and less arduous search journey. However, the question remains whether this enhanced usability necessarily equates to a superior or even responsible standard for the future of search.

Navigating the Ethical Labyrinth: Trust, Explainability, and Human Agency
While the efficiency gains are undeniable, conversational AI search engines introduce a complex set of ethical considerations, particularly regarding trust, explainability, and human agency. When using tools like Perplexity, users delegate a significant portion of their decision-making to the AI. The system selects, extracts, and summarizes content, crafting an answer based on its algorithmic assessment of relevance. Crucially, these tools often fall short in communicating why specific sources were chosen over others, neglecting a fundamental aspect of AI explainability (XAI). This lack of transparency directly impacts the trustworthiness of the system.
As Francesca Rossi, IBM’s Global Ethics Leader, emphasizes in "Building Trust in Artificial Intelligence," AI systems raise concerns about their ability to make fair decisions, align with human values, and explain their reasoning. The enterprise world is increasingly recognizing the risks associated with untrustworthy AI, including potential regulatory fines, reputational damage, and legal challenges. For individual users, often referred to as "Internauts" due to their highly skilled and habitual use of the internet, these risks might seem less immediate but are equally profound.
The architect Roberta Katz’s analogy—"first you make the building and then the building makes you"—aptly applies to digital systems. Just as physical environments shape behavior, the architecture of our apps and IT systems profoundly influences their users. It is a valid hypothesis that the convenience of a ready-made, authoritative-sounding answer from a tool like Perplexity could lead users to uncritically accept information, potentially diminishing their capacity to question accuracy or veracity. While AI search engines provide links to source documents, users still face challenges in quickly discerning:
- Which of the listed sources were genuinely used to formulate the answer.
- The relative weight or importance the AI assigned to each source.
While the first issue can also plague traditional search engines, the second stems from the deeper delegation of decision-making inherent in generative AI. Even with full explainability, the ease of access to summarized information could, in the long run, impact users’ critical thinking skills, much as text messaging has been linked to a decline in formal writing proficiency. The "tedious" process of sifting through Google’s SERP, comparing findings, and piecing together information from various articles, though more burdensome, constitutes a healthy exercise for analytical and creative skills. This process helps foster discernment and mitigates an uncritical over-reliance on AI models.
If the prevailing mental model of a search engine transforms into a tool that always provides the "right" answer, the implications for our ability to critically discern truth from falsehood could be immense. While this might be inconsequential for mundane queries (e.g., finding trousers or stain removal tips), it becomes critically important for knowledge acquisition in fields such as policymaking, scientific research, or civil society engagement.

The design of AI search experiences is not monolithic. Andi, for instance, offers an alternative architecture that aligns more closely with traditional search mental models, despite its chatbot-like interface. In its search results, the brief answer provided is often a direct snippet from a primary source like Wikipedia, rather than a generated summary. The results page also prominently displays a list of all sources with direct links, appearing on the left side of the page. Conceptually, this structure is not far removed from a traditional SERP, despite its distinct layout. This design choice addresses a critical concern raised by Kevin Roose of The New York Times: "If AI search engines can reliably summarize what’s happening in Gaza or tell users which toaster to buy, why would anyone visit a publisher’s Web site ever again?" Andi’s interface, by encouraging clicks to source websites, offers a partial solution, demonstrating the power of UX and UI design in fostering desirable user behaviors and sustaining the broader information ecosystem.
The Road Ahead: Building Trustworthy AI Search
The future of search will undoubtedly be shaped by AI. Generative capabilities offer myriad ways to enhance user experience and accelerate knowledge acquisition. However, the critical questions are not if AI will be the future of search, but how it will be implemented. Key inquiries revolve around the necessary guardrails and user interface design patterns, and how these tools can ensure trustworthiness. A study published in Nature underscored the conflicting views among users of AI science search engines, with some researchers praising their utility and accuracy, while others expressed concerns about inconsistency and a lack of trust in retrieval performance.
Given that trust is a central factor influencing AI adoption, two key areas of focus emerge for the development of future AI search engines:
- Technical Grounding and Accuracy:
- Retrieval-Augmented Generation (RAG): Integrating RAG architectures is crucial. RAG combines the generative power of LLMs with verifiable information retrieval, ensuring that answers are grounded in specific, attributable sources. This helps to mitigate "AI hallucinations," where models generate plausible but factually incorrect information.
- Enhanced Source Attribution: Beyond simply listing links, systems need to clearly indicate which parts of a synthesized answer come from which specific sources, providing granular transparency.
- Fact-Checking and Verification Layers: Incorporating automated or human-assisted fact-checking mechanisms can bolster the accuracy and reliability of AI-generated summaries.
- User Experience and Ethical Design:
- Explainable AI (XAI) in Practice: Designing interfaces that explicitly show how an answer was derived, including the relative importance of sources or the reasoning behind specific summaries. This could involve visual cues, interactive explanations, or confidence scores.
- User Control and Feedback: Empowering users to provide feedback on accuracy, suggest alternative sources, or adjust the level of summarization versus direct source access.
- Transparency on Limitations: Clearly communicating when an AI’s knowledge base might be limited, outdated, or when a query falls into a high-stakes domain requiring critical human review.
- Preserving Critical Engagement: Designing UIs that encourage, rather than bypass, critical thinking. This could involve presenting multiple perspectives, highlighting contentious points, or prompting users to delve deeper into original sources for complex topics.
- Adherence to Regulatory Frameworks: As global regulations like the EU AI Act emerge, AI search engines must comply with requirements for safety, transparency, and accountability, mitigating risks of bias and misinformation.
Within a context where users are increasingly delegating significant decision-making to LLMs, design efforts must prioritize accuracy, trustworthiness, and the comprehensiveness of search outputs. This extends far beyond merely building a user-friendly conversational interface; it necessitates a multidisciplinary approach that integrates ethical considerations, robust engineering, and human-centered design principles to ensure that the convenience of AI search does not come at the expense of informed discernment and a healthy information ecosystem. The ultimate goal is not just a good search experience, but the right experience—one that empowers users while upholding the integrity of knowledge.
