Why Content-led AI Matters in Higher Education

There is a growing body of evidence that unchecked AI systems have developed a habit,one that is quietly and consistently working against pedagogy:

They like to agree with you. A lot.

This phenomenon is called AI sycophancy–the tendency for AI models to prioritize user approval over factual accuracy, logical consistency or objective truth.

Recent research has found that some of the most widely used AI assistants agree with users at rates approaching 60% or higher (Fanous et al, 2025), often without the user realizing and regardless of whether the user's view is accurate.

It emerges from how these systems are trained: human evaluators, often without realising it, rate agreeable responses more favorably than accurate ones. The AI learns that agreement equals success. Over millions of training examples, at massive scale, that lesson becomes baked in.

For many use cases, this is a minor inconvenience. For higher education, it presents a fundamental problem.

What sycophancy looks like in a pedagogical context

Picture a student who has misunderstood a concept - a biological process, or a mathematical theory. They ask their AI assistant to explain it or check their understanding.

The AI, trained to be agreeable, responds in a way that subtly validates their misunderstanding, while appearing measured and authoritative. It picks up on the intent of their question and reflects it back to them, polished and confident.

Where an interaction with an instructor or professor would have corrected the misunderstanding, the AI has enabled the student to walk away with their inaccurate knowledge reinforced.

Research into AI and critical thinking in educational settings has highlighted exactly this risk: that overreliance on AI tools can hinder independent reasoning and reduce active engagement, particularly when AI fills the role of a knowledge authority rather than a learning scaffold (Aref, 2026).

When AI consistently tells students what they want to hear, it removes the productive friction that builds genuine understanding.

Now scale that across an institution. Across thousands of students, in hundreds of courses, over years of study. The pedagogical risk becomes significant. The question isn’t whether students will use AI - they already are, and increasingly so (HEPI, 2026).

The question is which AI mediates this risk and encourages constructive feedback, instead of reducing it.

Check your sources

The majority of AI draws from the open web, so it’s been trained on vast amounts of data - some accurate, some outdated, and some just plain wrong. When most AI generates a response, it lacks meaningful references, so there’s no guarantee that the response aligns with a tangible academic publication.

Even the well-established AI models oftentimes don’t understand the difference between what a faculty member has selected as required course material and a page that happens to rank highly on a search engine. They don’t know which chapters matter, which marking criteria are being utilised or what a student’s weekly learning objectives are.

That gap has potential to cause catastrophic implications on learning outcomes - and it’s the same gap that Kortext IQ was built to close.

Intelligence you can trust: how Kortext IQ works differently

As an Instructure Verified Integration partner, Kortext delivers world-leading digital teaching and learning solutions that are deeply embedded into Canvas LMS.

Our latest innovation provides a seamless answer to the sycophancy problem, delivering world-class AI optimized for higher education directly where students and their educators practice pedagogy – inside the LMS.

Kortext IQ is an AI assistant purpose-built for higher education. It’s not a generic chatbot being used in academic settings - it was designed from the ground up to operate within the specific context of teaching and learning.

The way it sources information is what sets it apart.

Kortext IQ was built and designed in answer to a challenge posed to us by Saïd Business School at the University of Oxford. As Canvas LMS users, Oxford Saïd needed an AI assistant that could:

• Aggregate, surface and augment content from across systems
• Provide value at the right time to the right user
• Support a global learner population across different time zones and languages
• Operate within secure, university-governed systems, services and frameworks
• Deliver hyper-personalized teaching & learning at scale in Canvas LMS

We built secure, deeply integrated enterprise-grade software delivering each of these objectives and more, providing exactly the carefully designed AI that higher education needs to encourage appropriate, academically aligned usage.

Kortext IQ draws exclusively from inside the guardrails of institution-approved content: the Kortext inventory of 4 million+ eBooks, faculty-assigned readings and LMS-integrated learning materials that have been provisioned by educators themselves and shared into the Kortext knowledge estate.

When a student asks a question, IQ draws its response from those approved sources, referencing every answer with a link to the relevant resource, making its reasoning transparent and protecting academic integrity.

This means two things:

1. IQ can’t tell students what they want to hear from the open web, because it’s not looking there. Its responses are grounded in the same materials the faculty member selected, meaning any answer it gives is one that can be traced, fact-checked and evaluated.
2. IQ operates across the whole text. Not just excerpts a student has read or the passages they were directed to. The full scope of the content is available to IQ when formulating a response.

This matters because it means a student who has only engaged with one section of a textbook cannot inadvertently narrow IQ's view of the material.

It also means that students can ask IQ questions about their own course design. For example,

“Please tell me my assigned reading for week 3 in course PSYCH675.”

Kortext IQ looks into your individual Canvas instance, referencing from your own assigned courses, uploaded files, journal articles, lecture notes and Kortext reading to answer questions specific to your personal academic context.

The honest design at the heart of IQ

Here’s something worth considering: the most effective AI for learning might not be the one that gives the best answers. It might be the one that prompts more questions.

Critical evaluation is a core component of higher education, the ability to take evidence and explore the ways in which it is concurrently accurate and limited. Kortext IQ encourages students to prompt thinking and evaluate evidence, rather than take it as fact.

Rather than simply providing a definitive response that a student can copy and paste, IQ is built to guide learners through university-approved materials, prompting reflection, surfacing further academic content and encouraging engagement with the text itself.

The Socratic Method has long been recognized as one of the most effective tools for developing critical thinking (Overholser, 1993): not telling people what to think but asking questions that require them to problem solve and deepen their engagement with material.

Kortext IQ brings that approach into a digital learning environment, ring-fenced around institution-approved academic content.

AI at the source of pedagogy: IQ and Canvas

Kortext IQ integrates directly with Canvas, this means students and academics encounter IQ at the center of course creation and independent study - not as an external tool, but as an embedded part of their Canvas experience.

For students, this means being able to engage with course content using AI support that’s grounded in the materials their institution has provisioned. They can explore texts, ask questions about course content and get support that’s aligned with what they’re learning.

It can also be accessed from directly inside their Kortext bookshelf, giving students a consistent and synchronised AI-enhanced reading experience.

For faculty, the benefits run deeper. Kortext IQ is not just a student-facing tool, it directly builds from Kortext fusion – Kortext's foundational AI-powered layer that aggregates data from systems across the institution.

This data gives faculty visibility into how students are engaging with content: which resources are being used or neglected, where students appear to be struggling and the implications of these on course design, decision-making and student success.

This closes a crucial gap: the gap between institution-provisioned resources and the learning outcomes those resources produce. IQ helps translate student engagement into tangible actions for improved pedagogical outcomes.

The AI concerns that can't be overlooked

Beyond pedagogy, institutions have significant and legitimate concerns about AI governance, asking questions like:

• How do you protect institutional data?
• How do you give institutions control over AI use?
• How do you support adoption?
• How do you ensure equitable access?
• How do you fit within our existing ecosystem?

At Kortext, our AI is designed to be embedded, not bolted on. It’s AI that’s been designed for institutional governance from the ground up.

Institutional leaders can define where IQ operates, what content it draws from and the parameters within which it works. For universities navigating the complex question of responsible AI adoption, this matters.

Kortext IQ offers a buy-not-build solution that delivers fast time-to-value, limitless scalability, predictable costs and governance frameworks that meet the stringent standards higher education requires.

Kortext IQ and Canvas: better together

The partnership between Kortext and Instructure represents something genuinely meaningful for higher education: a unification of Kortext’s transformative data, analytics, content and AI solutions with Instructure’s world-class learning management infrastructure.

Canvas is where teaching and learning happens for millions of students. Kortext brings to that environment a content layer - an inventory of 4 million+ eBooks - and an AI layer that knows how to work with them responsibly, pedagogically and in a way that’s honest and outcomes-focused.

Most AI in education is concerned about what it can produce. Kortext IQ focuses on what students can learn.

That distinction is what is currently missing in AI in higher education today. Not more shortcuts but genuine, purposeful engagement with material.

Together, that is what Kortext IQ and Canvas can deliver.

Kortext is going to InstructureCon 2026! Meet us at stand 52 and explore responsible, pedagogically aligned AI, directly integrated with your Canvas environment. Our InstructureCon session: Thursday 23 July, 10:55-11:15am, M112 Conference Theater

Learn more: https://www.kortext.com/us/events/instructurecon-2026