How to Create Ads With AI: A Step-by-Step Guide

To create ads with AI, give a tool your product photo and a preset, generate several variations, then score them before you spend. Modern AI ad generators handle the production — backgrounds, scenes, AI actors, voiceovers, and layouts — so the skill that matters is judgment, not design. The winning workflow is volume plus a filter: generate many angles cheaply, score them against proven performance patterns, fix or kill the weak ones, and put budget only behind the strongest.

For most of advertising history, the bottleneck was production. Making an ad meant a photographer, a designer, or a video editor, plus days of turnaround. AI collapsed that bottleneck to minutes — which sounds like the whole problem solved, until you realize the new bottleneck is knowing which of your hundred easy-to-make ads is actually worth running.

This guide walks through the full workflow, from a single product photo to scored, launch-ready creative.

Step 1: Start with your product and a goal

Begin with a clear product photo and one objective per ad — awareness, clicks, or conversions. A single clean image is usually enough for an AI generator to build from. Know who the ad is for and what outcome you're selling before you generate anything; the clearer the input, the better the output.

Step 2: Choose a format and preset

Decide whether you need image or video, and pick a style that matches the platform. Creator-style UGC video suits TikTok and Reels; clean product shots suit feed and shopping placements. Presets do the heavy lifting here — instead of prompting from scratch, you start from a proven creative template and adapt it.

Step 3: Generate several distinct variations

Don't generate one ad — generate several distinct concepts with different hooks, angles, and formats. The whole advantage of AI is cheap volume, which lets you explore angles you'd never pay to produce traditionally. Aim for genuine variety (different value propositions), not minor wording tweaks, so your test actually teaches you something about what resonates.

Step 4: Score before you spend

This is the step most people skip — and it's the most important. Before launching, score each ad against patterns from creatives that already performed. This catches the weak hooks, cluttered visuals, and unclear offers that doom most ads, so you fund only the survivors. Our guide to testing ad creative before you spend covers this in depth.

Step 5: Launch, learn, and refresh

Put budget behind the strongest creatives, let real data confirm the winners, then generate fresh variations of what works. This closes the loop and keeps you ahead of ad fatigue, since you always have new creative ready to rotate in.

Common pitfalls with AI ads

• Generating volume but never filtering it — spreading budget across weak creative.
• Producing minor variations instead of genuinely different angles.
• Ignoring platform-native format (running landscape video on vertical feeds).
• Letting AI choose the winner instead of scoring and testing.
• Forgetting disclosure rules for AI-generated spokespeople where required.

Create and score ads with AI in AdRoast

AdRoast is built around this exact workflow. Turn one product photo into image and video ads with 1,000+ presets, then roast each one against $25B of real ad spend for a score and plain-language feedback on the hook, visual, copy, and CTA — before you spend a dollar. You get AI's production speed and the judgment layer that turns volume into results. New to the space? Start with our comparison of the best AI ad generators.

Frequently asked questions