1. Home
2. Guides
3. How to make ads with AI

AI tools can produce an image or a video clip in seconds. A good ad still needs someone to decide who it is for, what it says and why anyone should care. This guide walks through a five-step workflow that takes one idea from a written brief to a finished ad, ready to post. It works with whichever image and video tools you prefer.

Step 1: Find the hook and write the brief

Most weak AI ads fail before the first prompt. The tool can only make what you describe, so the quality of the ad depends on how clearly you know what you want. Start with a one-page creative brief. It takes ten minutes and saves hours of random generating.

Answer these questions in plain sentences:

The hook is the angle that makes someone stop scrolling. Write three possible hooks before you commit to one. For the candle shop, that might be "The gift that outlives the candle", "Poured by hand, one jar at a time" or "Moving in? Make it smell like home." Pick the one that best matches the problem your audience actually has.

Step 2: Make the visual

Start from real photos of your product wherever you can. AI image tools are good at settings, light and mood. They are much less reliable at keeping the exact shape, colour and label of a specific product. Real reference photos keep the result honest and recognisable. Our guide to AI product photos covers how to shoot them on a phone.

Write prompts with the same structure every time, so you can change one thing and see what it does:

• Subject: the product and what it is doing
• Setting: where it is, and what is around it
• Light: soft window light, golden hour, studio softbox
• Camera: close-up, eye level, shallow depth of field
• Mood and style: calm, premium, playful, documentary
• Shape: the aspect ratio you need for the final ad

Example promptA soy candle in a clear glass jar on a light oak shelf beside a folded linen towel, soft morning window light from the left, close-up at eye level, shallow depth of field, calm and warm, neutral colours, vertical 9:16 composition with empty space at the top for a headline.

Generate several variations, pick the strongest and refine from there. Keep a simple prompt log (prompt, tool, settings, result) so you can recreate a look later. Leave headlines and logos out of the image itself: most image generators still render text and fine logo detail unreliably. You will add them cleanly in Step 4.

Step 3: Add the motion

Short video usually holds attention better than a still image on social feeds. Image-to-video tools can animate the visuals you made in Step 2, which keeps your video consistent with your stills.

• Plan the shots first. Three to five short shots are enough for most social ads. Write one line per shot: what we see, and how the camera moves.
• One move per clip. Ask for a single, clear action, such as a slow push in, steam rising or a hand lifting the jar. Busy prompts tend to produce warped, confusing clips.
• Keep clips short. A few seconds per shot is easier to generate cleanly and easier to edit.
• Make several takes. Generation is unpredictable. Make a few versions of each shot and keep the best.
• Check every frame. Look for melting hands, flickering edges, changing product shapes and warped labels. Cut anything that looks wrong.

Our guide to AI video ads goes deeper into shot planning, motion prompts and editing.

Step 4: Build the ad

Now combine your visuals, clips, headline and call to action in an editing or design tool. A simple structure works for most products:

1. Hook: the first moment shows the most striking visual or the hook line.
2. Product and benefit: what it is and why it matters to this audience.
3. Proof: a detail, a demonstration or a real photo that backs up the claim.
4. Call to action: one clear next step, with your logo.

Keep the headline short enough to read in a second or two. Add captions to any spoken words, because many people scroll with the sound off. Place text away from the screen edges, where platform buttons and captions can cover it.

Step 5: Prepare to publish

Each platform has preferred shapes. Export a version for each place the ad will run instead of letting the platform crop it.

Put together a simple publishing pack: each size of the ad, the caption text, hashtags if you use them, and the link. Make two or three variations with different hooks, so you can see which one performs best instead of guessing.

Advertising rules still apply

An AI-made ad is still an ad. The same rules on truthful claims apply as for any other advertising, wherever you run it. Examples include the CAP Code in the UK, the Singapore Code of Advertising Practice, the Malaysian Code of Advertising Practice, the FTC’s truth-in-advertising rules in the US and consumer protection law across the EU.

In practice, that means your AI visuals must not make a product look bigger, better or different from what customers will receive. Several platforms, including YouTube, TikTok and Meta, also have rules about labelling realistic AI-generated content. Policies change often, so check the current rules for each platform before you publish.

Common mistakes

• Prompting without a brief. You get attractive images that say nothing.
• Letting AI redraw the product. Customers notice when the jar, label or colour is wrong.
• Text baked into the image. Garbled lettering looks careless. Add text in your editor.
• Too many ideas in one ad. One message per ad. Make another ad for the next message.
• One size for every platform. Cropped headlines and logos waste the work.
• Skipping the frame-by-frame check. A single warped hand can undo a good ad.

More free guides

AI product photos that still look like your productShoot reference photos on your phone, then build campaign images without distorting the product.Read the guide →How to make AI video ads for a small businessPlan a 15-second ad, write motion prompts, edit, caption and export for Reels, TikTok and YouTube.Read the guide →