Comparing Elicit, ChatGPT Deep Research, and Kompas AI: UX, Capabilities & Use Cases

33 min readFeb 15, 2025

In the fast-evolving landscape of AI research assistants, tools like Elicit, ChatGPT’s Deep Research, and Kompas AI have emerged to help users sift through information and generate insights. Each tool offers a different user experience (UX), feature set, and ideal use cases. This article provides an in-depth comparison of these three platforms, evaluating their strengths and weaknesses in an objective manner. We’ll examine how each tool handles research tasks, the kind of output and UX they offer, and where they shine or fall short — all while subtly highlighting what Kompas AI brings to the table in terms of ongoing research, long-form report generation, and refinement.

Elicit: AI Research Assistant for Academic Literature

Elicit is an AI-powered research assistant designed primarily for academic use. Developed by Ought, it targets researchers, scientists, professors, and students looking for scientific information.

Elicit’s UX is more akin to a specialized search engine than a chat app. Users input a research question and receive a list of relevant academic papers from a database of over 125 million publications. Alongside the search results, Elicit provides one-sentence summaries of paper abstracts, key findings, and even the ability to extract specific data points into tables. In essence, Elicit feels like “Google Scholar meets ChatGPT,” offering the familiarity of academic search with the convenience of AI-driven summarization.

Strengths and Capabilities of Elicit

Academic Focus & Summarization: Elicit excels at literature reviews and academic deep dives. It uses advanced AI models (from the Allen Institute for AI) to understand queries and identify key findings across papers. It generates meaningful summaries, giving a concise answer to your question with direct links to relevant papers and highlighted annotations. This helps researchers quickly grasp the core ideas of multiple studies without reading each in full.
Intelligent Paper Discovery: The tool goes beyond keyword matching. Elicit’s AI-driven search can find relevant papers even if they don’t exactly match the query terms, by interpreting the context of the question. It also allows iterative searching — for instance, you can mark certain papers as relevant and ask for more like those, helping you dig deeper into a topic.
Data Extraction & Organization: Elicit can pull out specific details from papers (such as sample sizes, key results, or limitations) and organize them into a table or list. This feature is valuable for comparing findings across studies or gathering data for meta-analysis. It effectively automates parts of the systematic review process by extracting and aggregating information.
User Experience: The interface is structured and research-oriented. Rather than a free-form chat, you interact through structured queries and results. This can be efficient for users who prefer a more traditional research workflow. Elicit also integrates some reference management capabilities (e.g., you can save or “star” papers within the tool), making it easier to keep track of sources.

Limitations of Elicit

While powerful for academic research, Elicit has some limitations to be aware of:

Scope of Information: Elicit is heavily focused on scholarly literature. It doesn’t crawl the entire web for information; instead, it pulls from academic databases (like Semantic Scholar). This means it shines in empirical, data-rich domains, but might struggle with topics that aren’t well-covered in academic papers or that require theoretical discussion. For example, if you ask about a very recent news event or a niche cultural topic, Elicit may not find much, since those wouldn’t appear in academic journals.
Abstracts vs. Full Text: One noted limitation is that Elicit often only has access to the abstract or summary of many papers, rather than the full text. If a paper’s full content is behind a paywall or not open-access, Elicit can still tell you what the abstract says, but it cannot retrieve the entire paper. As a result, the answers it gives are sometimes based just on abstracts, which might omit important details or nuances. Users may need to obtain the full papers themselves for deeper analysis.
Learning Curve: For those used to simpler Q&A chatbots, Elicit’s interface might feel less straightforward initially. It offers a lot of options (filters for study type, the ability to add columns to result tables, etc.), which can be overwhelming until you get used to it. The learning curve could potentially hinder immediate productivity gains for new users as they adjust to the tool’s features. However, researchers often find that once they learn how to harness Elicit, it significantly speeds up their workflow.
Not a Writing Tool: Importantly, Elicit doesn’t generate long-form narratives or reports. It helps you find and summarize sources, but crafting a full report or paper is still up to the user. In other words, Elicit provides pieces of the puzzle (facts, summaries, data points), but you assemble those pieces into a coherent write-up yourself. This contrasts with more end-to-end tools like ChatGPT’s Deep Research or Kompas AI, which attempt to produce a polished report.

Best Use Cases for Elicit

Elicit’s strengths make it ideal for certain scenarios, especially in academia and data-driven fields:

Literature Reviews & Academic Research: Elicit is tailor-made for scanning academic literature. A graduate student or scholar can use it to quickly find the most relevant papers on a topic and get concise summaries of each. For example, if you’re researching a medical question, Elicit can surface recent studies and summarize their findings, saving hours of manual search. (Its developers claim up to 90% of Elicit’s information is accurate, reflecting its focus on scholarly sources.)
Evidence Gathering: If you need to gather evidence or examples from research (for instance, “What are all the documented effects of a certain drug across studies?”), Elicit can compile an answer by synthesizing many papers. This is useful for writing research-backed articles, policy briefs, or even business whitepapers that require citations from scientific literature.
Brainstorming Research Questions: Elicit includes features to help formulate research questions and even suggests related search terms you might not have considered. Researchers starting a new project can use it to explore a field, identify gaps in the literature, or refine their question by seeing what has already been studied.
Industry and Market Research: Beyond academia, Elicit can aid industry professionals who want academically grounded insights. For example, a market research analyst could use Elicit to find studies on consumer behavior or emerging technology trends. It can quickly pull out key findings from academic or industry papers about competitors or market trends, which can then be included in reports or strategy documents (with the caveat that the content is academic in nature).
Teaching and Learning: Professors or students might use Elicit to gather background readings or to ensure they cite relevant work. It’s also a way to double-check facts: asking Elicit a question is like querying a very smart academic search engine that not only finds sources but tells you what those sources say. This can augment learning by pointing to primary literature rather than just web summaries.

Summary: Elicit offers a specialized, research-centric UX that is highly effective for academic literature discovery and summarization. It is best for users who need to gather and synthesize scholarly information quickly. Its limitations in scope and full-text access mean it’s not a one-stop solution for all research needs — but within its domain, it’s a powerful assistant that significantly streamlines the research process. If your work revolves around papers and data, Elicit can save you time by automating tedious tasks (like skimming papers or extracting key points), allowing you to focus on analysis and interpretation.

ChatGPT’s Deep Research: AI-Powered Research Synthesis in a Chat UX

ChatGPT Deep Research is a feature of OpenAI’s ChatGPT (available to ChatGPT Pro users) that transforms the familiar chatbot into an AI research agent. Announced in late 2024, Deep Research is essentially an autonomous browsing and analysis mode within ChatGPT. It bridges the gap between a simple Q&A chatbot and a full-fledged research assistant. The user experience starts similarly to a normal ChatGPT interaction: you type in a prompt or question on a topic of interest. The key difference is clicking the “Deep Research” button, which triggers the agent to go beyond its built-in knowledge and actually search the internet for up-to-date information.

Once engaged, ChatGPT’s Deep Research will often ask you a few clarifying questions about your query to narrow or specify the scope. For example, if you request a report on a broad subject, it might ask whether you’re interested in a certain perspective or how technical the answer should be

platformer.news

. After you provide any needed clarification, the agent goes to work: it browses relevant websites, articles, and data sources, analyzes the content, and then synthesizes a comprehensive answer. The output is delivered as a long-form report or detailed response, complete with citations to the sources it used.

UX and Workflow

The Deep Research feature still lives within a chat-based interface, which makes it feel accessible and interactive. You see the agent’s questions as chat messages and you answer in natural language. During the research process, the interface might show that it’s browsing or thinking (and in some cases, it can display snippets of its chain-of-thought or the steps it’s taking). After a few minutes, it returns a thorough write-up. Users have reported that it can produce extensive reports — in one test, it generated a 4,955-word report in about five minutes. The report it delivers is typically well-structured: it might include an introduction, sections with subheadings for different aspects of the topic, and a conclusion or summary. Importantly, it cites its sources, often with footnotes or inline references, so you can verify the information provided.

From a user perspective, this is like having an on-demand research analyst. You ask a question, perhaps clarify the requirements, and then receive a written analysis as if a human had spent hours on it — all within a single chat conversation. The tone and depth of the output can often be adjusted by how you phrase your initial query or how you answer the follow-up questions (for instance, you can request a formal tone suitable for a report, or ask for an executive summary style, etc.). The convenience of this within a chat UX means you don’t have to switch contexts between different tools; it’s integrated into a platform many users already find familiar.

Capabilities and Strengths of ChatGPT Deep Research

Comprehensive Web Research: The chief capability of Deep Research is its ability to gather and synthesize information from across the web. It can read dozens of webpages, articles, and documents related to your query. OpenAI designed it to act like a “personal research assistant that can quickly become an expert on any topic”. In practice, you can ask it to research something and it will “read tons of websites, put all that data together, and give you back a report with proper sources — kind of like what you’d get from a professional researcher”. This means that rather than you manually Googling a topic and piecing together information, the AI does the heavy lifting.
Speed and Efficiency: For the level of detail it provides, Deep Research works remarkably fast. As noted, multi-thousand-word reports can be ready in minutes. Casey Newton, after hands-on testing, noted that Deep Research was “impressively competent” and can accelerate tasks like research, analysis, and argument formation in knowledge work. In other words, it’s a productivity boon — tasks that might take a human hours of reading and writing can be done in a fraction of the time. This speed is particularly valuable in professional settings where quick turnarounds are needed.
Quality of Analysis: Users have found that the quality of the analysis, on average, represents a step forward in AI capabilities
The Deep Research agent doesn’t just copy verbatim from sources; it summarizes and compares information, sometimes even drawing out insights or common themes from the data it gathers. For example, when comparing technologies or strategies, it might list pros and cons or highlight where multiple sources agree. One report compared Google’s own “deep research” prototype to ChatGPT’s output and found Google’s version to be much shorter and higher-level, whereas ChatGPT’s was far more detailed, exploring “nitty-gritty tradeoffs at great length”. This indicates that ChatGPT Deep Research tends to produce in-depth discussions rather than superficial answers.
Citation and Source Transparency: A significant strength for professional use is that Deep Research provides citations. It clearly references where information is coming from, which is critical if you need to verify facts or include the findings in a formal report. This feature turns ChatGPT’s output from a black-box answer into something you can fact-check and trust (with caution). Having sources listed also speeds up the transition from AI-generated draft to final product, since a user can easily incorporate the cited references or quote the original sources if needed.
Adaptability to Different Domains: Because it uses the breadth of the internet, ChatGPT Deep Research can handle a wide array of topics — from technical IT questions to market research to historical analysis. It can analyze any content that’s publicly available online or that you upload for it. This makes it a generalist tool suitable for many industries. Early impressions suggest it could fit into workflows for academics (to get quick summaries of a field), journalists (to gather background info), product managers (to survey market needs or tech trends), or policy analysts
. Essentially, anyone who benefits from synthesizing large amounts of information could leverage this tool.

Limitations and Considerations for ChatGPT Deep Research

Despite its powerful capabilities, ChatGPT’s Deep Research has notable limitations and is not without flaws:

Access and Cost: Currently, Deep Research is only available to subscribers of ChatGPT’s highest tier (Pro at $200 per month). Moreover, usage is metered — users are limited to 100 deep research queries a month. This high cost and limit mean the feature is geared towards enterprise users or professionals for whom the productivity gains justify the expense. For casual users or those on a budget, this is a significant barrier. The pricing suggests that running these extensive research tasks is compute-intensive (indeed, OpenAI notes the feature is resource-heavy, which is why it’s limited). Thus, organizations might reserve Deep Research for high-value queries.
Time and Interaction: Each deep research query can take several minutes to complete. OpenAI indicated tasks might take 5 to 30 minutes or more depending on complexity. This isn’t the instant answer you get from normal ChatGPT. While it’s impressive given the work done, users must be prepared to wait for results. Additionally, the need to answer clarifying questions means it’s not entirely hands-off; you might need a back-and-forth at the start to set the right direction. If the query is misinterpreted or too broad, you might have to refine and run it again, consuming one of your limited queries.
Accuracy and Hallucinations: Like any large language model, ChatGPT can still hallucinate information or make incorrect inferences. OpenAI warns that the research agent may mix in some false facts or fail to discern fact from fiction, especially given the open web’s content. It might also inadvertently include information from less reliable sources if they were part of the web results. For example, it could pick up an erroneous detail from a random blog or an outdated statistic. While the inclusion of citations helps catch these (you can check if a claim is actually supported by the source), it means the output isn’t guaranteed to be 100% correct or trustworthy without verification. This is a crucial point: the AI’s synthesis is only as good as the sources it finds and its ability to interpret them correctly. Users should review the output and not treat it as gospel.
Scope Limitations: Currently, Deep Research can only access publicly available information on the web (plus any documents you manually upload for it to analyze). It cannot penetrate paywalled content or databases that require subscriptions (e.g., certain academic journals, proprietary industry reports). It’s also limited to what’s been digitized; if information hasn’t been put online, the agent won’t find it. In an enterprise context, it doesn’t automatically have access to internal documents or private data unless you feed them into it. OpenAI is exploring connecting it to more specialized or private data sources in the future, but those are not in place yet. For now, this means some research tasks will still require traditional methods or additional tools if they involve non-public data.
Formatting and Editing: Users have noted that the initial output sometimes has minor formatting quirks or citation formatting issues (since the agent is essentially writing a report on the fly). While it generally produces a coherent structure, you might need to do some cleanup or reformatting if you’re going to use the text in a publication or slide deck. Also, because the output comes as a single large chat response, editing it might involve copying it out to a word processor. Unlike Kompas AI (discussed later), ChatGPT doesn’t offer a dedicated report editing interface for refining the result — it’s up to the user to prompt further or do manual edits outside the chat.
Ethical and Confidentiality Concerns: In professional settings, one must be cautious about the data sent to an AI. While ChatGPT can analyze uploaded files, companies might be hesitant to upload sensitive reports or data for analysis due to confidentiality. OpenAI has policies and now allows opting out of data being used for training, but still, sending proprietary info to an external AI service can be a concern. Thus, for enterprise use, Deep Research might be limited to public or non-sensitive queries unless proper assurances are in place.

Use Cases for ChatGPT Deep Research

ChatGPT’s Deep Research mode is versatile, but certain use cases particularly play to its strengths:

Quick Turnaround Research Reports: In scenarios where a comprehensive brief or report is needed on short notice, Deep Research is a game-changer. For instance, a business leader could ask for an analysis of a new market development or a competitor’s strategy, and get a detailed report with citations within the hour. In one case, the tool generated a thorough report on how publishers can benefit from the Fediverse with real-world examples and analysis. Tasks that would normally require a researcher or analyst working for days can be at least initially drafted by Deep Research, then reviewed by humans.
Professional Writing & Analysis: Consultants, analysts, or journalists can use Deep Research to gather background information and different viewpoints on a topic. For example, a policy analyst could prompt it to gather international case studies on a policy issue, or a journalist might use it to compile facts and quotes from various news sources about an event. The tool’s ability to incorporate multiple perspectives and examples (with references) is especially useful here. Axios described ChatGPT’s Deep Research as a “promising intern” — it can do a first pass on research, which the professional can then refine. In enterprise environments, it can function as a tireless research assistant for staff, boosting productivity.
Learning and Exploration: For individual power users, Deep Research is a great way to learn about a new field or topic thoroughly. If you want to become conversant in, say, the latest developments in renewable energy or an overview of machine learning techniques in finance, you can ask for a deep research report. The result will serve as a curated summary of the topic, often highlighting key concepts, trends, and references to follow up on. It’s like getting a crash course from an expert who also hands you a reading list (via the citations).
Enterprise Knowledge Synthesis: Companies can leverage Deep Research to compile internal knowledge. For example, if a corporation has a bunch of whitepapers, reports, or wiki pages, feeding them into ChatGPT (via uploads) and asking it to synthesize could produce useful internal briefs. While this strays into knowledge management territory, the ability to quickly summarize large volumes of text (even if user-provided) with an AI could save employees time. This is especially relevant when onboarding new team members or when decision-makers need a summary of past research.
Comparative Analysis and Decision Support: When faced with multiple options or complex decisions, Deep Research can compare and contrast information. A product manager could ask, “Compare the features, pricing, and user feedback of software X vs software Y vs software Z in a report,” and the agent will try to gather web reviews, specification sheets, etc., and lay out the differences. Having an AI-generated comparison can surface important factors and save preliminary research time. Similarly, for scientific questions, one could ask for a summary of evidence for and against a certain hypothesis based on published studies.

Summary: ChatGPT’s Deep Research mode is a powerful tool for broad, fast research synthesis, embedded in a conversational interface. Its ability to produce a well-structured, cited report on almost any topic with minimal user effort is indeed a “game-changer” for many professionals. The UX is engaging and relatively straightforward — you converse with it like you would with ChatGPT, but get a far more extensive output. However, its high cost and current limitations mean it’s best suited for users who truly need that level of depth regularly (e.g., analysts, researchers, consultants in enterprise settings). It’s an excellent “first draft” generator and research aid, but usually requires a savvy user to guide it initially and to verify and refine the results. When used appropriately, it can dramatically reduce the time spent on gathering information and allow users to focus on higher-level analysis and decision-making.

Kompas AI: Iterative Deep Research and Long-Form Report Generation

Kompas AI is a relatively new entrant that positions itself as a comprehensive platform for multi-step research and report writing. In contrast to ChatGPT’s chat-focused approach or Elicit’s paper summaries, Kompas AI offers a structured, report-ready UX from the outset. It was built with the goal of automating in-depth research across many sources and delivering the findings in a coherent long-form document. In essence, Kompas AI acts like an AI research analyst and a report writer combined, with a focus on ongoing, iterative investigation and easy refinement of results.

Kompas AI’s workflow is designed to handle research as a multi-stage process rather than a single query-response. When you start with Kompas, you typically provide a research topic or a broad question. The system automatically plans the research process, often by breaking the topic into an outline of subtopics or key questions (essentially creating a research plan). It then deploys its AI engines to gather information on each part of that outline. According to the documentation, Kompas AI “iteratively analyzes hundreds of pages” to provide truly comprehensive insights. It pulls from a broad range of reliable online sources and filters out extraneous or irrelevant data in the process, so the user isn’t overwhelmed with noise.

Strengths and Capabilities of Kompas AI

Multi-Depth Iterative Research: One of Kompas AI’s standout features is its iterative approach. Rather than answering a query in one shot, it allows for continuous deepening of the research. After the initial results are compiled, you can use the “Research Further” option to drill down even more on any aspect. This means if one round of research brings up new questions or areas that need more detail, Kompas can automatically go back out and fetch additional information. The process is continuous until you are satisfied with the depth of coverage. This is particularly useful for complex topics where new sub-questions arise during exploration. It ensures the final output isn’t just a surface-level summary, but can be expanded step by step into a truly in-depth study.
Structured Report Generation: Kompas AI is built around producing a long-form report as the end product. It doesn’t just dump a long answer; it organizes findings into a clear, structured format. For example, if researching a market trend, it might automatically generate sections like Introduction, Market Overview, Key Trends, Case Studies, Challenges, Conclusion, etc., populating each with synthesized insights from sources. This structured approach makes it easy to either skim for main points or dive into details as needed. The output is essentially report-ready — with logical flow and segmentation — which saves users a lot of time in editing and organizing content.
In-Depth Analysis and Synthesis: Beyond just gathering facts, Kompas performs analysis on the collected data. It uses advanced analytics to identify pertinent trends, correlations, and conclusions across sources. In other words, it doesn’t treat each source in isolation; it looks at the body of evidence and draws higher-level insights. For instance, if multiple sources agree on a point, the report will highlight that consensus; if there are differing opinions or data points, it can compare them. Kompas also evaluates the reliability of sources before incorporating them, which helps ensure the report’s content is credible and not based on fringe or dubious information. By the time the information appears in your report, it’s been through a synthesis process aimed at accuracy and coherence.
User Experience & Editing: Kompas AI offers a user-friendly, document-centric interface. Instead of a plain chat window, users interact with a workspace where the emerging report is visible and editable. Notably, it includes an AI Edit feature that can automatically adjust the tone of the report, translate content, or refine phrasing as needed. This is helpful for tailoring the final document to a specific audience or style (e.g., making the language more formal for an official report, or simpler for a general audience). Additionally, users can manually edit the report at any time; you have full control to rewrite or tweak any section. The combination of AI-driven drafting with human-in-the-loop editing provides a high degree of refinement and polish. In practice, this feels like using a smart word processor that can not only correct grammar but also fetch new content on command.
Seamless Integration & Sharing: After completing research, Kompas makes it easy to share the results. You can export the report as a PDF or Word document with just a few clicks, meaning the output can directly be used as a deliverable. This is a convenience factor: rather than copying text from a chat and formatting it in a separate document, Kompas gives you a ready-to-go report. Additionally, Kompas AI is cloud-based and works across devices (web browser on Mac/Windows, and even mobile) with synchronization. Professionals on the go can start research on one device and continue on another. The platform also emphasizes privacy and security (data encryption, not using your content for training, etc.), which can be crucial for enterprise adoption.
Broad Topic Coverage with Depth: Kompas handles a wide variety of research topics, from market trend analysis to competitor overviews to exploratory topic research. It’s versatile in application — business strategy teams, academic researchers, and content writers can all use it. The difference is that Kompas is particularly suited for cases where you need a comprehensive document at the end. For example, it’s excellent for writing detailed reports, whitepapers, case studies, or technical analyses where multiple sources need to be consulted and the findings compiled. It’s like having a research department that gathers facts, and a writing team that drafts the report, all in one tool. This end-to-end capability (from gathering to final write-up) is where Kompas AI differentiates itself the most.
Comparison to ChatGPT & Others: To put Kompas’s approach in context, it explicitly aims to go beyond the quick Q&A style of a typical chatbot. As the FAQ notes, “While ChatGPT focuses on quick question-and-answer interactions, Kompas AI specializes in multi-step, in-depth research and long-form report creation.”. This design philosophy means Kompas may not be as immediately conversational as ChatGPT, but it digs deeper by default. Compared to a tool like Perplexity (another AI search engine), Kompas “goes a step further by facilitating multi-layered research and extensive document creation”. In short, Kompas is built for thoroughness and completeness of research, delivering a ready report rather than just an answer snippet or a chat exchange.

Use Cases for Kompas AI

Kompas AI’s strengths lend themselves to various scenarios, especially where a polished document is the desired outcome:

Market Research Reports: Suppose a company wants a report on emerging trends in electric vehicles or an overview of the AI startup landscape. Kompas AI can handle the breadth of such a task by collecting data on the market size, key players, consumer trends, recent developments, etc., and then generate a cohesive report. The user can iteratively refine the focus — for instance, after an initial draft, ask Kompas to “research further” into battery technology advancements or regulatory changes, adding those details to the report. The result is a comprehensive market research document that could be client-ready with minimal editing.
Competitor Analysis & Business Intelligence: Business strategists can use Kompas to automate competitor research. By inputting a prompt like “Analyze Company X’s recent strategies and compare them to Company Y,” Kompas will gather information from news articles, financial reports, interviews, and any reliable web sources. It will filter out irrelevant data (e.g., press fluff) and present key metrics and qualitative insights side by side. The final report might include sections on product offerings, market positioning, strengths/weaknesses, and recent moves of each competitor. This type of structured competitive overview is incredibly useful for strategy meetings or SWOT analyses.
Technical Documentation and Whitepapers: For professionals needing to create technical reports or whitepapers (e.g., an overview of cybersecurity threats in 2025, or a whitepaper on the benefits of a new software architecture), Kompas AI provides the research backbone. It can retrieve documentation, expert blog posts, academic papers, and case studies on the topic and compile the relevant findings. Because it evaluates the credibility of sources, it tends to include authoritative content (like well-known tech blogs or research papers). The writer can then use the AI Edit feature to ensure the tone is consistent and the jargon level is appropriate for the intended audience, yielding a publishable whitepaper draft much faster than starting from scratch.
Academic Research and Theses: While Elicit is more directly aimed at academia, Kompas can also support scholars, especially in the writing phase. Graduate students working on a thesis could employ Kompas to gather a wide range of references and facts for a literature review chapter, for example. The student can then refine the report, insert their own analysis, and use the compiled draft as a foundation. Kompas’s continuous research feature means as new papers or data come out (or if the student finds a gap in the argument), they can update the research incrementally. The tool’s ability to handle citations and present information in sections can help in organizing a thesis or dissertation content logically.
Policy and Case Study Reports: In government or NGOs, when preparing policy briefs or case study reports, having current and comprehensive information is key. Kompas AI can assist by pulling together case studies from various countries or regions, compiling statistics from reports, and summarizing expert analyses on the policy issue at hand. For example, a policy researcher could ask for a report on the impact of telemedicine in rural healthcare, and Kompas would gather data from healthcare studies, news, and pilot program results, presenting them in a structured way (perhaps broken down by benefits, challenges, examples, etc.). The researcher can then fine-tune the narrative and ensure it aligns with their policy recommendations. The structured output saves a huge amount of time in collating information and lets the expert focus on interpretation.
Continuous Research Projects: Some projects aren’t one-time queries but require ongoing research updates — for instance, a team tracking the development of a technology or monitoring news on a particular topic over months. Kompas’s iterative nature is well-suited for this. One can maintain a Kompas project and periodically hit “Research Further” to fetch the latest information and append or update the report. This way, the report becomes a living document that grows over time. Teams can collaboratively refine it, and Kompas ensures new data is assimilated. Such a use case is valuable for, say, an investment firm keeping an active research dossier on an industry, or a product team staying updated on emerging user feedback and research in their field.

Summary: Kompas AI stands out as a tool purpose-built for deep, iterative research and the creation of long-form, structured outputs. Its UX is geared towards producing a tangible report rather than just an answer, which differentiates it from the chat-based format of many AI assistants. By guiding the user through an outline, gathering vast information, and facilitating easy refinement, Kompas can deliver comprehensive reports that are essentially ready to share. In doing so, it addresses a pain point that professionals often face: not just finding information, but organizing it into a usable document. Kompas AI’s strengths lie in its thoroughness and its focus on refinement — you don’t just get an answer, you get a draft report that you can iteratively improve. For users who regularly need to produce research-backed reports or need an AI partner for extended investigative work, Kompas AI provides a robust solution that combines the roles of researcher and report writer in one platform.

Comparative Analysis: How Elicit, ChatGPT Deep Research, and Kompas AI Stack Up

Each of these three AI tools serves the core goal of assisting with research, but they do so in distinct ways. The best choice often depends on the context — the kind of information you need, the format you want it in, and the workflow you prefer. Below is a structured comparison of key aspects of Elicit, ChatGPT’s Deep Research, and Kompas AI:

User Experience (UX) and Interface

Elicit: Offers a search-based interface. The UX feels similar to using an academic search engine or database. You pose a query and get a list of papers and summaries. It’s menu-driven and table-oriented — great for systematically scanning results. However, it’s not conversational. You interact by refining search terms or selecting filters, which is comfortable for researchers but less so for general users expecting a chat. Elicit’s interface also provides tools like tables and saved lists which are powerful for analysis but require learning to use effectively.
ChatGPT Deep Research: Lives in a conversational UI (the ChatGPT chat). This makes it very approachable — you can type naturally, and the AI will handle the complexity behind the scenes. The experience is interactive: the AI might ask questions back, which feels like talking to a colleague or an assistant. There’s little to no setup; if you can chat, you can use it. The downside is that the final output appears as a long message in the chat, which you might need to copy elsewhere for heavy editing or formatting. There’s also not much visual structure in the interface beyond the text it outputs (no built-in outline view or section collapsing, for instance). It’s simplicity vs. structure — the chat UX is simple, but not purpose-built for presenting lengthy research findings.
Kompas AI: Provides a structured workspace geared towards report writing. The UX is more like a document editor combined with a research dashboard. You see an outline or sections being built out, and you can navigate through different parts of the research easily. This structured format makes it clear where everything will go in the final report. For users who think in terms of documents, this feels intuitive — you see the report taking shape as the AI gathers info. It might be a bit more involved than a plain chat (there are likely buttons or options to “research further,” edit, etc.), but it is designed so that even non-technical users can follow along, because the information is organized logically. In short, Kompas’s UX is report-centric and “ready-to-use” — the tool itself helps you get to a final polished document with minimal fuss in external editors.

Capabilities and Depth of Information

Elicit: Specializes in academic and scholarly content. Its capability is narrow but deep in that domain. It can quickly surface peer-reviewed studies, complete with summaries and extracted data. However, it doesn’t autonomously browse news sites, blogs, or general web content. If the information you need resides in academic papers, Elicit’s AI search and summarization is extremely powerful. It also handles multiple papers at once (for instance, summarizing four papers side-by-side, or chatting with up to four papers simultaneously in its premium version). But ask Elicit about a non-academic topic (say, “opinions on the best travel destinations in 2024”) and it won’t be helpful — that’s not its arena.
ChatGPT Deep Research: Has a broad scope — effectively anything on the internet is fair game. It can gather news articles, forum discussions, educational content, statistics from reports, etc. The depth it reaches depends on what’s available online and how it interprets your query. It’s very good at giving a big-picture synthesis because it’s not limited to one type of source. However, because it covers breadth, it may not always dive as deep into niche academic details as a specialized tool like Elicit would. It might get the general gist from a few papers but not enumerate every experimental detail, for example, unless prompted. That said, with the right guidance, it can also incorporate scholarly content (it could find open-access papers or summaries). Another capability is that it can analyze any text you feed it, which means you can give it company reports, datasets (in text form), or other materials and have those included in the research — a flexibility the other two tools lack.
Kompas AI: Aims for both breadth and depth by structuring the research process. It automatically spans multiple layers of inquiry — first giving broad coverage, then drilling down where needed. This means its capability is not limited to one domain; it will use news, academic sources, industry sites, etc., whatever is relevant and credible for the topic. Kompas’s iterative approach ensures that if a subject needs more granular detail, the tool can focus on that specifically in a follow-up round. For example, if researching a medical policy, it might pull high-level stats from WHO in the first round, and then in a second iteration grab details from specific case studies or local data. The result is a capability to cover a topic comprehensively — from overview down to fine details. Additionally, because Kompas evaluates source reliability, it tries to maintain a high quality of information, combining the credibility focus of Elicit with the wide net of ChatGPT’s web browsing. It may not match Elicit in, say, certain nuanced academic tasks (like extracting a specific figure from a paywalled paper), nor can it tap into internal databases without input, but for public information it covers a lot of ground confidently.

Output Format and Refinement

Elicit: Output is provided in pieces — short answers, lists of papers, tables of extracted data. It’s then up to the user to incorporate these into a final document. Elicit doesn’t directly give you a narrative or a report. On the plus side, the facts it provides are easy to transfer into your work, and you can trust that they are linked to sources (since it’s literally giving you the papers). However, the burden of writing lies on you. If you need a well-written summary, you’ll have to compose it from Elicit’s notes or use another writing tool. Elicit is excellent for gathering raw material and insights, but not for producing the polished prose around those insights.
ChatGPT Deep Research: Output is a detailed written report or essay in the chat. This is great because you get narrative, explanations, comparisons — essentially a draft that can often read quite well. The language is usually polished (ChatGPT is known for fluent writing), and it attempts to be comprehensive. You can even ask it to format in markdown or include bullet points if you want a certain style. However, because it’s one large block of text (in most cases), refining it can be a bit awkward in the chat interface. You might ask follow-up questions in chat to expand or clarify certain sections, but major restructuring or edits likely require copying the text out. There’s no direct feature to rearrange sections or easily tweak the tone within ChatGPT beyond more prompts. So, the output format is immediate and verbose, but further refinement is a manual or external process.
Kompas AI: Output is a fully structured report with clear sections, which you can refine both with AI assistance and manually. This is arguably Kompas’s forte — the report is not just a blob of text; it’s segmented, titled, and can even include things like an automatic table of contents or references section. Because the tool itself supports editing, you can polish the document in situ. For instance, if a paragraph seems off, you can rewrite it directly or ask the AI to adjust it. If you want to insert a company-specific insight or confidential data point, you can do that seamlessly in the draft. Essentially, Kompas provides a report editing environment where the AI-generated content is the starting point, and you have Word-processor-like control to finalize it. The final document can be exported cleanly as a PDF or DOCX, preserving the formatting. This end-to-end handling of output — from generation to refinement to export — is where Kompas really differentiates itself. It’s built for producing deliverables, not just raw answers.

Strengths Summary

Elicit’s Strengths: Extremely useful for academic research and data-driven inquiries. It’s precise in fetching scholarly evidence and summarizing it. Great for literature reviews, finding citations, and extracting specific info. Its focus means it’s less likely to introduce non-validated info. Also, it’s affordable (a robust free version and a ~$10/month plan), making it accessible to students and researchers on a budget.
ChatGPT Deep Research’s Strengths: Unmatched in versatility and convenience for broad research tasks. It combines a powerful language model with live web data, delivering coherent narratives. It’s like having a competent generalist researcher who can summarize the internet for you. The conversational aspect means even non-researchers can use it easily — just ask for what you need. It’s a huge time-saver for creating first drafts of reports, and the clear citing of sources lends credibility to its outputs.
Kompas AI’s Strengths: Excels in depth and deliverable-quality output. Its iterative digging ensures no stone is left unturned if you need thoroughness. It effectively project-manages the research for you: outlining, gathering, analyzing, and compiling, all within one platform. The result is a well-structured report that can be readily used in a professional context. Kompas also supports refinement tools (both AI-driven edits and manual control) which means the output can reach a high level of polish without leaving the platform. This makes it ideal for ongoing research projects and situations where the final format is as important as the content (e.g., client reports, publications).

Weaknesses and Caveats Summary

Elicit’s Weaknesses: Limited to academic content — not helpful for general web info or creative tasks. Sometimes only provides abstracts, which might miss details. It doesn’t generate full narrative answers, so it’s not a one-stop solution for writing. The interface, while powerful, can be a bit complex for new users, and maximizing its utility may require some learning and effort.
ChatGPT Deep Research’s Weaknesses: High cost and restricted access put it out of reach for many users (especially individuals or small orgs). It can still make mistakes or include dubious info if not carefully monitored, so you can’t blindly trust everything it outputs. It’s also dependent on internet sources — if information is scarce or too new, the agent might struggle. The need for verification remains, meaning professional users must fact-check the AI’s work. Additionally, working with the output beyond reading it (like repurposing it in reports) requires extra steps, as the chat interface isn’t designed for heavy editing or formatting tasks.
Kompas AI’s Weaknesses: As a newer specialized tool, it may not have the same level of community or third-party integration as more established platforms. Users might find that very simple queries are faster to just do in a normal search or with a quick chatbot; Kompas’s process might be overkill if you only needed a straightforward fact or a one-paragraph answer. In other words, its thorough, multi-step approach, while a strength for big research, could be less convenient for trivial questions or extremely time-sensitive quick answers. Also, because Kompas tries to be comprehensive, there’s a chance it could include more information than needed, requiring the user to prune the report to keep it focused (though this is easier to do in Kompas than in a chat). Finally, Kompas likely has its own subscription model (it offers free options with possibly paid tiers for more usage), so users will have to consider yet another platform in their toolkit — adopting it means adjusting to a new workflow, which some may resist if they are already comfortable with existing tools.

Choosing the Right Tool for the Job

To sum up the comparison, here’s when you might choose each tool:

Choose Elicit if you are conducting academic research, literature reviews, or need quick access to scientific evidence. It’s perfect for when your question is, “What do academic studies say about X?” For a student writing a research paper or a scientist scanning the latest publications, Elicit provides a focused lens into scholarly knowledge. Its ability to pull out key findings from papers is like having a research assistant who instantly reads dozens of PDFs for you. However, if your research extends beyond academia or if you need a compiled report, you’ll need to use Elicit in conjunction with your own writing and perhaps other tools.
Choose ChatGPT Deep Research if you require a broad synthesis of information across various source types and you want it quickly and in a readable form. This is the go-to for one-off comprehensive briefs — for instance, a startup founder preparing for a meeting on a topic outside their expertise could get up to speed with a Deep Research report. It’s also suitable in enterprise settings where the cost is justified: e.g., consulting firms or media companies that will heavily utilize its capabilities for client work or content creation. If you value having a conversation with the AI to iteratively shape the output, ChatGPT provides that interactive experience. Just be ready to double-check the facts it gives and budget the queries wisely.
Choose Kompas AI if your end goal is a well-structured, in-depth report and you anticipate an iterative research process. It’s the ideal choice for thorough investigations where you want to ensure completeness and accuracy at each step. For example, if you’re tasked with writing a detailed industry report quarterly, Kompas can help you build and update that report systematically. It’s also excellent for collaborative environments — a team can use the tool to accumulate research over time and produce a consistent report that multiple stakeholders can edit. Kompas AI particularly shines in professional and enterprise scenarios where the presentation of research is as important as the findings themselves. Its report-centric approach and refinement tools mean the gap from research to deliverable is very small. If you prefer a more guided experience that leads to a tangible product (as opposed to just an answer), Kompas is a strong choice.

Ultimately, these tools aren’t mutually exclusive. A savvy researcher might use Elicit to gather academic references on a topic, then use ChatGPT’s Deep Research to get a broad narrative including news and web commentary, and finally use Kompas AI to organize everything into a formal report with their own insights added. Each tool brings something unique: Elicit the precision, ChatGPT Deep Research the breadth and narrative, and Kompas AI the structure and depth.

Conclusion

The advent of AI research assistants has dramatically expanded what individuals and teams can do when it comes to gathering and synthesizing information. Elicit, ChatGPT’s Deep Research, and Kompas AI represent three different approaches on this spectrum, each with a focus that makes it particularly well-suited to certain users and tasks.

In an objective comparison, we saw that:

Elicit offers a scholar’s lens — it’s exceptional for quickly mining academic knowledge and is best harnessed by those who need credible, citable findings from scientific literature. Its user experience prioritizes efficient discovery and extraction of data, though it stops short of writing the report for you.
ChatGPT Deep Research acts like a Swiss army knife for research, handling an array of information and delivering human-like analysis in record time. It’s a powerful aide for professionals who need comprehensive answers and can invest in this premium service, but it requires careful use and verification.
Kompas AI, while less of a household name, emerges as a robust end-to-end research and reporting tool. It subtly outperforms in scenarios requiring iterative deep dives and polished output, thanks to its multi-step research strategy and integrated writing environment. Kompas AI’s strengths in continuous research, long-form report generation, and user-guided refinement make it a compelling choice for ongoing projects and complex research needs.

The user experience (UX) varies accordingly — from Elicit’s targeted interface, to ChatGPT’s conversational ease, to Kompas’s structured workspace. A tech professional might appreciate how Kompas AI structures findings in a report-ready format, or how Elicit provides quick access to hard data, whereas a general user might lean towards the approachable chat experience of ChatGPT for broad questions.

For those in academia or data-centric fields, Elicit can be a trusty sidekick. For those who need an all-around researcher on call (and are willing to pay for it), ChatGPT’s Deep Research is like hiring an intern with an encyclopedia of knowledge. And for those who need to produce detailed, refined reports regularly, Kompas AI quietly offers an all-in-one solution that handles everything from discovery to write-up.

In practice, the choice may come down to the specific use case and workflow preference. It’s even plausible to use these tools in combination, leveraging each of their strengths at different stages of a project. What’s clear is that AI is transforming research tasks: what used to require teams of people and weeks of effort can now be accomplished faster and, in some cases, by a single person with the right AI toolkit.

As with any AI tool, the key is understanding their outputs and limitations. None of these tools completely removes the need for human judgement — whether it’s interpreting results, verifying facts, or crafting insights. But they significantly augment our capabilities. By comparing Elicit, ChatGPT Deep Research, and Kompas AI, we gain insight into how AI can cater to different research needs: from quick academic queries, to broad knowledge synthesis, to comprehensive report creation. Armed with this knowledge, users can choose the tool (or combination of tools) that best fits their requirements, and confidently navigate the ever-growing sea of information with an AI compass to guide the way.