AI Tools For Learning Compared: ChatGPT, Gemini, Perplexity, NotebookLM and Learn Place

Millions of people now use AI tools to learn. They open ChatGPT, Gemini, Perplexity or NotebookLM, type a question, receive an answer, and believe they have learned something. Most have not. They have consumed information. Consumption and comprehension are not the same thing, and no amount of AI-generated text will close the gap between the two.

The distinction matters because the tool you choose to learn with determines the nature of the learning itself. A tool designed for information retrieval will produce information consumers. A tool designed around the psychology of how humans actually form understanding will produce people who understand. The market is saturated with the former. There is, at present, one example of the latter.

The Problem With Using General AI For Learning

ChatGPT, Gemini and Perplexity are general-purpose AI systems. They were not built to teach. They were built to respond. The difference is fundamental. A system built to respond optimises for the quality and speed of its output. A system built to teach must optimise for the transformation of the person receiving that output. These are opposing objectives.

When a general-purpose AI answers your question, it has no model of what you already know. It has no awareness of the gaps in your understanding. It cannot connect new information to your existing mental framework because it does not know what that framework contains. It simply generates the most probable useful response to the prompt it received. The result is information that may be accurate but is pedagogically inert.

This is not a criticism of these tools. They are extraordinary at what they were designed for. But learning is not what they were designed for, and using them as primary learning instruments is like using a calculator to learn mathematics. The output is correct. The understanding is absent.

ChatGPT: Articulate But Aimless

ChatGPT is the most widely used AI tool for learning, and in many ways the least suited to it. It is a conversational machine optimised to produce helpful, coherent responses. It will explain any concept you ask about. It will do so clearly, at length, and with impressive fluency. It will also explain it to you in exactly the same way whether you are a complete beginner or an expert, unless you explicitly tell it otherwise.

The learner bears the entire burden of directing the interaction. They must know what to ask, how to ask it, and when to move on. They must self-assess their own understanding and decide whether the explanation was sufficient. This is precisely the set of skills that a person who needs to learn a subject does not yet possess. The person who can perfectly direct their own learning through ChatGPT prompts is the person who least needs ChatGPT to learn.

ChatGPT has no concept of a learning journey. Each conversation is stateless in its pedagogical effect. It does not track what you have covered, what you struggled with, or what concept you need next. It is an oracle that answers whatever is placed before it. Oracles are useful. They are not teachers.

Gemini: Power Without Pedagogy

Google's Gemini brings multimodal capability and deep integration with the Google ecosystem. It can process documents, images, and code. It can draw on the breadth of Google's indexed knowledge. For research and productivity, it is formidable.

For learning, it shares every limitation of ChatGPT and adds one of its own. Gemini's strength in information synthesis means it tends to produce comprehensive, dense responses. Comprehensiveness is the enemy of learning. A learner does not need everything about a topic. They need the right thing, at the right time, connected to what they already know. Gemini cannot provide this because, like all general-purpose AI, it has no model of the learner.

The tool excels at answering the question "What is X?" It cannot answer the question "What does this person need to understand next, given what they already know and how they came to know it?" That second question is the only one that matters in education.

Perplexity: Research Disguised As Learning

Perplexity occupies a distinct position. It is an AI-powered research tool that retrieves, synthesises and cites information from across the web. For factual queries, source verification, and rapid literature review, it is arguably the best tool available. It does what traditional search engines should have done years ago.

The danger of Perplexity as a learning tool is precisely its competence at research. It makes the acquisition of information so frictionless that the user conflates finding an answer with understanding a concept. A person who uses Perplexity to research a topic will emerge with a well-sourced summary. They will not emerge with understanding, because understanding requires struggle, application, and the integration of new knowledge into existing cognitive structures. Perplexity removes the friction that learning requires.

Research is not learning. The person who has read every paper on a subject is not the same as the person who has worked in that subject. Perplexity is an exceptional research assistant. It is not a learning tool.

NotebookLM: Organisation Without Transformation

Google's NotebookLM is the closest any general tool comes to being learning-adjacent. It allows users to upload their own sources, documents, notes, PDFs, and then interact with an AI grounded in those materials. It can summarise, cross-reference, and generate study aids from the user's own content.

This is genuinely useful for organising existing knowledge. A student revising from lecture notes or a professional synthesising research papers will find real value in NotebookLM. But organising information is not the same as transforming the person who encounters it.

NotebookLM does not know whether you understood the material it summarised. It does not adapt its explanations to your prior knowledge. It does not challenge you to apply what you have read. It is a sophisticated filing cabinet that can talk. The information goes in, gets rearranged, and comes back out in a more digestible form. What it does not do is ensure that you have digested it.

The fundamental issue is the same as with every tool on this list. NotebookLM has no theory of learning. It has no model of how human understanding actually forms. It is, like the others, an information tool adopted for a purpose it was never built to serve.

The Missing Foundation: Apperception

Every tool discussed above fails at learning for the same reason. None of them are built on a theory of how humans learn. They are built on a theory of how language models generate text. These are entirely different problems, and solving the second does not solve the first.

Human understanding does not form through the passive reception of information, no matter how well presented. It forms through apperception, the process by which new experience is assimilated into, and transforms, the existing body of knowledge a person already holds. A new concept is not understood until it is connected to what came before it, until it reshapes the learner's prior mental model and becomes part of the framework through which they interpret subsequent experience.

This is not a new insight. It is a foundational principle of psychology that has been understood for over a century. Yet not a single mainstream AI tool applies it. ChatGPT, Gemini, Perplexity and NotebookLM all treat the learner as a blank surface onto which information is projected. They do not enquire about what the learner already knows. They do not structure their output to connect with existing understanding. They do not verify that apperception has occurred.

The result is millions of people who have consumed enormous quantities of AI-generated explanation and retained almost none of it. The information was never anchored. It was never connected to a lived framework of understanding. It arrived, was read, and evaporated.

Learn Place: Built On How Humans Actually Learn

Learn Place is the only AI learning platform built from the ground up on the psychological principles of apperception. It is not a general-purpose chatbot repurposed for education. It is a learning system that uses AI as the mechanism through which established principles of human understanding are applied at scale.

The difference is architectural, not cosmetic. When a learner engages with Learn Place, the system builds a model of their existing knowledge. It identifies what they already understand and what gaps exist. It then structures each explanation to connect new information to what the learner already knows, ensuring that apperception occurs rather than hoping that it might.

Before a single lesson begins, Learn Place generates a dynamic custom syllabus personalised to the individual. Not a static curriculum designed for the average student, but a learning path shaped by the learner's motivations, their current level of experience, and what they are trying to achieve. A career changer learning Python to move into data science receives a fundamentally different syllabus from a computer science graduate looking to deepen their understanding of distributed systems. The motivations are different. The starting points are different. The path through the material must be different.

This is where every other tool fails before it even begins. ChatGPT does not ask why you want to learn something. Gemini does not ask what you already know. Perplexity does not care whether you are a beginner or an expert. They all deliver the same information to every person, regardless of context. The learner is left to filter, prioritise and sequence the material themselves, a task that requires the very expertise they do not yet have.

The custom syllabus eliminates the single greatest source of wasted time in self-directed learning: studying the wrong things in the wrong order. A person who already understands variables and control flow does not need to sit through an explanation of them. A person motivated by building web applications does not need to start with theory that will not be relevant for months. Learn Place identifies what can be skipped, what needs reinforcement, and what sequence will produce understanding fastest, given who the learner actually is.

Every interaction is part of this structured learning journey, not an isolated conversation. The system tracks what has been covered, what was struggled with, and what the learner needs to encounter next. It does not wait to be asked the right question. It knows what question needs to be asked and when.

This is the difference between an AI that responds and an AI that teaches. A tool that responds relies on the learner to drive the process. A tool that teaches drives the process itself, adapting in real time to the learner's evolving understanding.

The Philosophy Of Doing

Apperception is the foundation. But understanding, genuine understanding, requires a second ingredient that no general-purpose AI provides. It requires doing.

A person can read every explanation of a concept and still not understand it. Understanding crystallises through application, through the act of using knowledge under real conditions, confronting the friction of practice, and emerging with the kind of comprehension that can only come from having done the work.

Learn Place embeds this philosophy into its core through Verified Experience. Learners do not simply consume explanations. They complete real tasks that require them to apply what they have learned. Their work is then verified to confirm that genuine understanding has formed, not just the ability to reproduce an AI-generated answer.

This is the critical gap that no other tool addresses. ChatGPT will explain recursion. Gemini will explain it with diagrams. Perplexity will cite the academic papers that defined it. NotebookLM will summarise your lecture notes on it. None of them will make you write a recursive function, watch you struggle through it, and verify that you understood why your first attempt failed.

Learn Place does exactly this. Understanding is not measured by what a person can repeat. It is measured by what they can do. Any model of learning that does not culminate in verified action is incomplete.

The Comparison That Matters

The question is not which AI tool produces the best explanation. They all produce good explanations. The question is which tool produces genuine understanding in the person who uses it.

ChatGPT, Gemini, Perplexity and NotebookLM are information tools. They are exceptional at what they do. But what they do is not learning. They retrieve, synthesise, and present information. The transformation of that information into understanding is left entirely to the learner, with no structure, no verification, and no psychological foundation.

Learn Place is a learning tool. It is built on the principle that understanding forms through the connection of new knowledge to existing knowledge, and that it is verified only through the application of that knowledge to real problems. It is the only tool in this comparison that was designed from first principles around how humans actually learn, rather than how language models generate text.

The people who use general AI tools to learn will know a great deal. The people who use Learn Place will understand what they know, and be able to prove it. In a world where information is free and abundant, the ability to demonstrate genuine understanding is the only thing that still has value.