AI Transparency

Last updated: March 2026

Fiveable helps high school students and teachers prepare for AP exams. From 2018 to 2023, all content was created and edited by humans. New AI models in the last few years have changed the way we work and created new opportunities for how we can support students and teachers.

This page explains where AI shows up on Fiveable, how our content is created, and the principles that guide our work.

Our Principles

We believe AI has enormous potential to expand access to high-quality educational resources and help all learners understand rigorous content. These are the principles that guide every decision we make about AI use at Fiveable:

1. Grounded in real curriculum.

AI doesn't decide what's required knowledge. For AP courses, every piece of content is built from College Board learning objectives, skills, topics, and exam frameworks. For non-AP courses, open-source textbooks and established curricula set the same foundation. Without knowing exactly what's required, there's no way to know if what AI produces is right. The curriculum is what makes that possible. AI lets us create more from that source material (more question variations, more practice, more ways to engage with the same concepts), but the curriculum always sets the boundaries.

2. Our expertise directs the AI.

It's super challenging to get AI to produce high quality content, but we have unlocked it because of how we apply our expertise. Our team brings years of teaching experience, our founder was an AP reader, and we have nearly a decade of building content at Fiveable. All of that goes into how we prompt, build workflows, and evaluate. The original Fiveable study guides were written by hundreds of students and teachers, and the instincts from that work still shape everything we build today. Rather than every student and teacher trying to get solid output from ChatGPT on their own, we've done that work and baked it into the product.

3. Evaluations are required.

Everything we produce goes through rigorous evaluations. We check content across multiple dimensions: accuracy, curriculum alignment, difficulty, and quality. FRQ scoring is even further benchmarked against official released materials so we could tune the workflows until they score within half a point of readers. Different subjects often require different approaches, and we treat them that way. The bar is high because students shouldn't have to question whether what they're studying is correct. With that said, mistakes and errors can still happen, as they did with human writers too. Any content error can be reported and we apply fixes immediately.

4. Scarce resources should become abundant ones.

Practice questions, essay prompts, and quality study materials have historically been hard to come by, and the best ones are usually expensive. That means the students who need the most practice often have the least access to it. AI changes that equation. It lets us build resource libraries large enough that every student can get the repetition and variety they need to actually learn, not just review. To us, that's one of the most important problems AI can help solve in education and why we also value quantity. There should be so many practice prompts that teachers and students never have to ration.

5. New capabilities mean better content.

Our team tests new models, tools, and techniques as soon as they're available. When a new model meaningfully improves quality, we don't just apply it going forward. We go back and rebuild. Entire question sets have been regenerated because the technology became good enough to do them right. We run new evaluations, revisit older material as our systems improve, and update content as standards evolve. The goal is a library that keeps getting more accurate, more comprehensive, and more useful over time.

Where Students and Teachers Interact with AI Directly

These are features where AI is working in real time based on your input:

  • FRQ Grading. Write a response to an FRQ and get detailed feedback showing exactly which rubric points you earned and what to improve. Scoring is based on College Board-aligned rubrics and built with multi-step workflows, not a single prompt. (practice FRQs | teacher grading)

  • Visual Cheatsheets. AI-generated visual summaries of key concepts from topic study guides, designed for quick review. (see cheatsheets)

  • Quiz Me and Image Generation. Generate on-demand practice questions for any topic or create visual study aids from content in your study guides. Highlight any content in study guides to use these features.

  • Content Error Reports. Flag a potential error on any page and AI analyzes the report with full context from the original content generation. Clear issues are fixed automatically. If it's unclear what's wrong, the report gets routed to our team for review.

How Our Content Is Created

Not all content on Fiveable is created the same way. Here's how each type works:

AP Study Guides were originally written by hundreds of students and teachers over the years. We've since used AI to improve them, aligning content to what College Board actually tests, catching inaccuracies, and filling gaps. The base layer is human-written and AI has made it more accurate and comprehensive.

Non-AP Study Guides are AI-generated from open-source textbooks and established curricula, with the same structured workflows and quality checks as our AP content.

Practice Questions and FRQs are AI-generated using multi-step workflows. Each generation involves significant steering, context building, and evaluation against real exam standards.

Data and Privacy

When students use AI-powered features on Fiveable, we send only the data necessary to provide the feature (for example, an essay for grading or a topic for question generation). We do not use student data to train AI models. For full details, see our Privacy Policy.

Limitations

AI is not perfect. Content can have errors and scores are estimates, not official grades. If you spot an issue, you can report it directly from any page on Fiveable and we'll fix it.

That said, the rate of innovation from Anthropic, Google, and OpenAI has been truly remarkable. There are far fewer hallucinations and errors now than there were six months ago, and even fewer than a year ago. We fully believe that in the next six months and the next year, it will continue to get better. And as it does, so will everything on Fiveable.

Questions?

If you have questions about how Fiveable uses AI, reach out to us at help@fiveable.me.

2,589 studying โ†’