I Tested 8 AI Presentation Tools

October 8, 2026

This past weekend I road-tested eight AI presentation products and recorded the whole thing.

Please don't watch it. It's really boring. Go to the bottom to see the prompt I used, the resulting decks, and the final rankings.

I don't think the results and rankings are anywhere near as important as what this whole exercise tells me about where AI presentation is heading.

The last time I sat down to try multiple tools was maybe a year and a half ago or longer and I was really surprised by how intense and excellent the competition is.

Here's a collage of the designs

You genuinely have an embarrassing riches of options:

I expect the field to look completely different again in six months.

People have said generative AI is powerful but has no taste. That's changing. Browse Twitter and you'll see tons of UI components, shaders, and motion graphics. A lot of them really are beautiful.

In this round, Gamma 5 on Max Effort took design taste to another level, and Bolt and Moda were a pleasant surprise too.

Recent usage research from Menlo Ventures, Andreessen Horowitz, and others shows that power users drive a disproportionate share of spend on generative AI. I think a class of power users who value taste and design will pay for better design.

Pulling in deep research and your own context to make a professional-looking deck is now table stakes, and those decks are starting to look the same.

The next level is bespoke visuals for every deck, even every occasion.

Here, for instance. The background art is fully custom - aligned with the deck's theme and the slide's content.

And this is a custom HTML generated on the spot to convey a concept. Also fully aligned with the colors and style of the deck's theme.

Before generative AI, making custom graphics or video for every talk cost too much or wasn't possible. Now you can make custom images, graphics, motion graphics, and HTML from a simple prompt, fast enough that if you give seven talks in a week, each one can have its own assets.

Give it reference images of your product, a place, or your team, lock in a style, and it can redraw all of them to match that talk.

If a hard idea needs a moving flow chart, build it for that one talk. So what if you never use it again?

Make it easy for people to use all their context to build the visual story they want.

The irony is that this round, I typed one prompt into each tool, because I don't have connectors into any of them. Next time, I hope to test through integrations like MCP or a command-line tool, so I can feed in my own materials, preferences, and skills and see how well each tool does.

That's clearly where this is going. You tell the agent: "pull these ten sources from my files, check them against ten others, my audience is this group, the talk is twenty minutes, and I want twenty slides. It builds a deck for that talk and only that talk. Before, that made no sense. You reused one deck for a hundred talks because nobody has time to build a hundred decks. Now it's possible, and soon it'll be easy.

I use one single prompt and ran that generation across all of them.

Please note that I didn't use a precise scoring system and calculate the final ranking to decimal points. It's more like I assigned grades on each of these aspects but then the final ranking is basically me asking myself, "After all that how would I rank them?"