I've been building presentation software for a while. Before that, I used AI tools to create decks for my own projects. And I've watched teammates, friends, and strangers on the internet stumble through the same patterns over and over. So here are the five mistakes I keep seeing — and what I do instead.
1. Treating AI like a mindless ghostwriter
The most common mistake is handing the AI a vague topic and expecting magic. "Write me a presentation about quarterly results." You get back generic slides filled with fluff. Bullet points that say "focus on growth" and "improve efficiency." Nothing wrong with those words — but they don't mean anything to your specific situation.
What works for me: I give the AI a skeleton first. I write down three or four rough slide titles, plus a sentence for each about what I actually want to say. Then I ask the AI to flesh it out. The output is much more focused because I've done the thinking upfront. The AI is good at phrasing, not at knowing what matters to my team.
My teammate told me last week he spent two hours rewriting slides that the AI generated from a one-line prompt. I've been there. Feed the AI better inputs.
2. Trusting the data without checking
AI models are great at sounding confident. They'll generate statistics, cite studies, even invent references that look real. I once asked an AI to list market share numbers for presentation software. It gave me neat figures — all wrong. The sources were plausible but nonexistent.
This is dangerous in a business presentation. Someone in the room will fact-check you. Or worse, you'll present to a client and they'll notice the numbers don't add up. I always treat AI-generated data as a starting point, not a final answer. I verify key numbers myself or at minimum ask the AI to show its reasoning. Some tools now include citation features. Use them, but still glance at the original source.
3. Ignoring the visual logic
People copy-paste AI text into slides and call it a day. The result is walls of text. The AI doesn't know your slide size, your audience, or how much time you have. It writes paragraphs. A good presentation isn't a document on a screen — it's a support for spoken words.
Here's a rule I try to follow: each slide should make one point. Not three, not a list of seven items. One. If the AI gives me a list of five bullet points, I pick the strongest one and turn the rest into separate slides or drop them. While building Huanjian (huanjianai.tech) we tested this a lot — we found that the best results come from telling the AI to "make each slide contain no more than 20 words" and then editing manually. The visual design does the rest.
4. Over-relying on AI for the narrative arc
AI can help you structure a presentation, but it doesn't know your audience's inside jokes, the tension in the room, or the question someone asked in the hallway before the meeting. I've seen people let the AI define the entire flow, and the result feels generic. It starts with "the current landscape" and ends with "next steps." That's fine for a template. It's not fine for a presentation that needs to persuade a skeptical VP.
I still sketch the narrative on paper first. Usually just a few Post-its: problem, solution, proof, call to action. Then I ask the AI to help me flesh out each part. The AI suggests transitions and examples. But the emotional arc — the reason someone should care — that's mine to decide.
5. Forgetting that you are the slide deck
The last mistake is the opposite of the first. Some people hand everything over to the AI, including the speaker notes. They read from the script the AI generated. The result is a monotone delivery with sentences that don't sound like them. I've done this myself. I recorded a practice run and cringed at how robotic I sounded.
Your slides are a tool, not a script. The AI can give you bullet points for reference, but you have to translate them into your own voice. When I prepare, I rewrite the speaker notes as if I'm explaining the idea to a friend over coffee. Short words. Pauses. No jargon unless it's necessary. The AI gives me raw material — I turn it into something that sounds like me.
None of this is revolutionary. It's just the stuff I wish someone had told me earlier. I still mess up sometimes. But avoiding these five mistakes has saved me dozens of hours and a few embarrassing meetings. Try one next time you open an AI tool for slides. See if it makes the output feel less like a robot wrote it.