When Gemini Slides Go Wrong: A Better Workflow for Building Lesson Slides

By Ben | Taught AI

One issue I’m seeing more and more with schools using Gemini is this:

A teacher creates a strong lesson, clicks into Canvas to turn it into slides, then tries to edit them… and suddenly everything becomes a wall of text, broken formatting, or code-like confusion.

In some cases, it even feels like the work has been lost.

The key point is this: this isn’t user error. It’s a reflection of early-stage technology that is still evolving.

Gemini is incredibly powerful, but slide generation and editing are not yet as stable as people expect. Right now, the best results come from treating Canvas as the final step, not the main workspace.

A Better Way to Work (For Now)

The most reliable approach is simple:

• Build and refine your lesson in text first

• Stay in chat mode and improve the content until it’s classroom-ready

• Only move to Canvas once you’re happy

• Then generate the slides in one go

• Finally, download and make any detailed edits yourself

Quick Example Workflow

• Export your Scheme of Work to Google Drive

• Upload it into the Taught AI Lesson Generator

• Plan and refine the lesson in text mode

• When ready, use Canvas to convert it into slides

Why This Matters

When users move too quickly into slide generation, they often lose control of the content. That’s when things start to feel messy or unreliable.

For now, the golden rule is:

Refine first. Convert last.

Final Thought

This will improve — quickly. But for now, schools need workflows that are stable, predictable, and low-risk.

The goal isn’t just faster slides. It’s better teaching materials staff can trust.