Are you spending hours every time creating lecture materials? You think about the content, add diagrams, adjust font sizes, and before you know it, it's night.
When I gave the same prompt to ChatGPT and Claude, ChatGPT won on content, while Claude won on visuals.
However, Claude's first draft wasn't ready to be projected directly in a lecture, and ChatGPT showed similar weaknesses when opened in my environment.
Two AI Slide Makers Launched Within a Month
The trigger was this statement by Boris Cherny, developer of Claude Code.
https://x.com/bcherny/status/2111158659971699136
Claude Slides was released on September 16. It is available on all plans including the free tier (Enterprise is in beta and requires owner activation).
ChatGPT Slides was released on October 16. It is available to Pro, Business, and Enterprise users who can use Spaces (shared workspaces). It can generate slides from conversations, files, or outlines, and allows design matching with sample slides.
Both can export to PowerPoint, PDF, and Google Slides. The ChatGPT version also supports simultaneous team editing and comments.
Generated Using the Same Prompt
The subject was a 20-minute beginner workshop. Here is the prompt given to both:
1Create slides for a 20-minute beginner workshop titled "How Does Changing Your Requests to AI Affect Work Results?" Include comparisons between bad and good requests, diagrams, and exercises for attendees to do on the spot. Keep it to 10-12 slides.
For Claude, I added this follow-up instruction:
1Please revise the exercise page so that attendees can fill it out on the spot.
I created the ChatGPT Slides myself.
On the Claude side, instead of using the chat interface at claude.ai, I used Claude Code (Opus 5.5) to generate it in the Claude Slides format.
Comparison Results
ChatGPT won on structure. With 12 slides, it included comparisons of bad and good requests using meeting notifications as examples, and pages displaying AI response examples for each.
The flow covered the four elements of a request (purpose, materials, conditions, final form), specific adaptation to work contexts, how to ask for revisions, verification before use, and next steps.
Each page had time estimates that summed up to exactly 20 minutes. Claude also included a schedule, but only ChatGPT displayed per-slide time estimates on the screen.
The most useful feature was the request templates.
They included instructions like "Do not add facts not present in the materials" and "If insufficient to create, please ask questions first." This was content ready to distribute during the lecture.
The exercises were two-tiered: rewriting the request, sending it to AI, and comparing results. Designed to run in three 1-minute cycles.
Claude won on visuals. With 11 slides, it featured a navy and orange color scheme and card-based layouts.
Comparisons included thank-you emails and document summaries. The flow covered the four elements of a good request (purpose, audience, materials, format), reusable request patterns, adding one sentence, exercises, rewrite examples, and three things to use starting tomorrow.
All slides included speaker notes, which is helpful for the presenter.
ChatGPT's design was primarily white background with blue text and thin lines, featuring only about two diagrams. Side-by-side, Claude was visually more striking.
I could only verify the revision of the exercise page on the Claude side.
After revision, it became a worksheet with four input fields and a "Completed Request" field arranged together. Attendees could fill them out from top to bottom to create a complete request.
Despite winning on visuals, weaknesses were clear. Claude had small text and many pages with empty lower halves. One heading had a single character drop to the next line.
On the ChatGPT side, when received in PowerPoint format and opened in my environment (LibreOffice), there were also pages with small text and noticeable blank spaces at the bottom. Since display varies due to font differences, I cannot definitively call this a flaw of ChatGPT.
My Verdict
Use ChatGPT if you want to use the content as-is; use Claude if you need presentation-ready visuals. That settles it.
ChatGPT excelled at designing the lecture structure itself. It anticipated instructor needs, filling in exercise procedures and template warnings.
Claude is strong in visual appeal when projected and provides excellent speaker notes. Claude Slides itself is available from the free plan (this version was created via Claude Code).
I would solidify the content skeleton with ChatGPT and finish the presentation materials with Claude.
The Unexpected Pitfall: "The First Draft"
What came out in the first draft was not ready for projection, at least for Claude.
Both had small text and blank spaces at the bottom of pages. While the ChatGPT result was viewed in my specific environment, the Claude issue was distinct.
The danger is that they look like proper slides at first glance.
The colors, cards, and even speaker notes are neatly organized, making you want to project them immediately.
But if you project them as-is, attendees will see tiny text and empty space.
Even with Claude, which won on visuals, my initial reaction was "Is this really okay?" Not a compliment. This is separate from who won.
Summary
Content for lectures: ChatGPT. Visuals for presentations: Claude.
However, in my hands, neither was ready for immediate projection after the first generation.
AI-generated slides are most dangerous right after they appear.
How far can you delegate lecture or internal training material creation to AI? Let's build a workflow tailored to your current methods.
Free consultation here:
