8 Tips for AI Flyers That Don’t Look Like AI Made Them

AI image tools have gotten better at handling text and layout, but most AI-generated flyers are still easy to spot: too busy, too many fonts, and imagery that feels slightly wrong. Writing for ZDNET on August 11, 2026, Elyse Betters Picaro shares eight tips for getting cleaner results from AI design tools, based on her own testing and a background in journalism, publishing, and art direction. The core finding: prompt quality and upfront constraints make more difference than which tool you pick.

What happened

ZDNET contributor Elyse Betters Picaro tested AI flyer generation and documented eight tips for getting usable results. She brings relevant credentials: a BA in journalism, an MFA in publishing, and experience as an art director for a university print magazine. Her starting observation is blunt. AI flyers are usually recognizable at a glance because of overcrowding, competing typefaces, and imagery that looks almost-but-not-quite real.

She tested the approach using her own 20-year high school reunion as a real brief, iterating on prompts to see what improved results. The article covers the full process from concept to export.

The 8 tips, in order

1. Talk to a graphic designer first. Even a brief conversation with a designer friend or a hire from a platform like Fiverr can catch problems before you distribute anything widely. Picaro notes that skilled designers bring accessibility knowledge and originality that AI cannot replicate.
2. Define the message before you touch any tool. Reduce what the flyer needs to say to one sentence. Adobe’s own design advice suggests establishing a “3-second message” so the headline and main visual do the heavy lifting. A useful framing: “This design is for __ and it should get people to __.”
3. Use a reference design for analysis, not copying. Find a professional flyer you like on Facebook, Etsy, or elsewhere. Upload it to ChatGPT or Gemini and ask it to break down the color palette, spacing, mood, typography, and information hierarchy. Use that analysis as input for your own prompt, but avoid commands like “copy this style exactly.” If you do not own the uploaded art, stick to public domain references. Picaro also recommends using Temporary Chat features in ChatGPT or Gemini so uploads are not saved or used for model training.
4. Specify format and impose hard limits. Before generating, decide where the flyer will appear. A vertical 4:5 Facebook image has very different constraints from an 8.5-by-11-inch print piece. Tell the model the exact dimensions, and set explicit limits on the number of fonts, the amount of text, and the number of visual elements.
5. Give the AI a clear creative direction. Vague prompts produce vague outputs. Describe the mood, the audience, and the visual tone in concrete terms rather than abstract ones.
6. Limit text in the image. AI models still struggle with typography. The less text you ask the model to render inside the image, the fewer errors you will get. Where possible, plan to add text in a separate tool after generating the base image.
7. Iterate with specific feedback. Rather than regenerating from scratch when something is off, tell the model exactly what to change. Name the element, describe the problem, and specify what you want instead.
8. Check for the classic AI tells before sharing. Scan for garbled text, extra fingers, mismatched proportions, and fonts that shift mid-word. These are the details that signal AI generation to an informed viewer.

Why it matters

Most small businesses and community organizers are not going to hire a designer for every event flyer. AI tools lower the barrier, but the default output is often worse than a basic Canva template because there are no guardrails. The tips here are really about adding those guardrails manually through better prompting.

For businesses running social media campaigns, flyer quality is not a vanity concern. A cluttered or visually off image signals low effort and can reduce engagement before anyone reads the copy. If you are also running paid placements, a weak creative will hurt your cost-per-click regardless of how well the targeting is set up.

The reference-design analysis step is the most underused idea here. Asking an AI to describe what makes a good design before asking it to produce one is a way of building a brief collaboratively rather than just throwing a vague request at the model and hoping for the best. We covered a similar principle in our look at how structured AI workflows produce more consistent outputs in coding contexts. The same logic applies to creative tasks.

Our take

The honest answer for most small businesses: use Canva with a paid template and spend 20 minutes customising it. You will get a better result faster than most AI prompting workflows currently allow, and you will not need to check every letter for hallucinated text.

That said, the reference-analysis approach in tip 3 is genuinely clever. Using the AI as a design critic before using it as a designer is a workflow worth testing. It forces you to articulate what good looks like before you start generating, which is the real problem with most AI creative briefs.

The tip about limiting text inside the generated image is critical and often ignored. AI text rendering is still unreliable. Generate the visual background in the AI tool, then layer your actual event details in Canva, Adobe Express, or even PowerPoint. This splits the job between what AI does well (visual composition) and what it still fumbles (accurate letterforms).

If your business regularly needs event promotion, social graphics, or print materials, it is worth exploring professional UI and design support rather than treating AI-generated output as finished work.

What to do about it

1. Write your one-sentence message before opening any tool.
2. Find a reference flyer you like and upload it to ChatGPT or Gemini (using Temporary Chat) to get a structured description of its design elements.
3. Build your prompt using that description, plus your message, your format dimensions, and explicit limits on fonts and text quantity.
4. Generate the background or visual composition only, then add all text in a separate tool like Canva or Adobe Express.
5. Before sharing, check the output for garbled text, font inconsistency, and any other obvious AI tells.

The best AI flyer is one where you cannot tell it was made with AI. That still takes more work than most people expect.

Source: ZDNET · AI