Plan a Facebook Ad Campaign With AI Before You Spend Anything

The expensive mistakes in Meta ads usually happen before the first ad goes live: the wrong objective, an audience of 4,000 people, a $5 budget split across six ad sets. Half an hour of planning catches most of them. AdChat can do that planning with you using real numbers from Meta, and because it only reads, you can do it before you've turned on write access.

At the end you'll have a one-page plan you can build from yourself, or hand to the AI with campaign from scratch.

Who this is for

Anyone about to launch a campaign, whether it's your first ever or your first for a new product.

What you need

• AdChat connected to your AI app or the built-in chat.
• No write access needed. Everything in this guide reads data or asks Meta for estimates.
• The basics of your offer: what you sell, roughly who buys it, where they live, and what you can afford to spend per month.

Step 1: Tell the AI what you're working with

Give it the whole picture in one message. The more specific, the better the plan.

I sell handmade leather dog collars for $45 from my Shopify store. Most buyers are women 30 to 55 in the US who treat their dogs like family. I can spend about $600 a month on ads. I want online sales. Help me plan a Facebook and Instagram campaign. Don't create anything yet.

If you already run ads, add:

Look at my ad account first and tell me what's worked before.

AdChat will pull your past results with get_insights and use them to ground the plan.

Step 2: Pick the objective

Which campaign objective should I use and why? What would I lose by choosing traffic instead of sales?

What a good answer looks like: a clear pick with a reason. For online sales, that's usually the Sales objective, with a note that it needs the Meta Pixel set up and recording purchases. If your Pixel isn't set up yet, the AI should tell you, because that changes the plan.

Step 3: Size the audience

Suggest three audience options: one broad, one built on interests, and one narrower. Estimate the size of each for the US, ages 30 to 55.

The AI uses search_interests to find Meta's interest IDs and estimate_audience_size to get a range for each option. You'll get something like:

• Broad, women 30 to 55, US: very large
• Interests in dog training and pet supplies: a few million
• Interests in a specific breed plus luxury goods: a few hundred thousand

Ask which one it would start with at your budget, and why. With a small budget, one broad or medium audience usually learns faster than three small ones. For more on this step, see audience size estimation.

Step 4: Set the budget and structure

With $600 a month, how should I set this up? How many ad sets, how much per day, and should the budget sit on the campaign or the ad set?

What a good answer looks like: a simple structure that fits the money. At $20 a day, that's usually one campaign, one or two ad sets and a few ads, not five ad sets at $4 each. Meta needs enough results in each ad set to learn who to show ads to, and splitting a small budget too thin slows that down.

Push back if the plan looks busy:

That's a lot of ad sets for $20 a day. Would you simplify it?

Step 5: Sketch the ads

Give me three different angles for the ads. For each, write a primary text, a headline, and describe the image or video I should use.

Ask for angles that are actually different: one on quality, one on the dog, one on the gift. Three versions of the same idea won't teach you much.

If you've run ads before:

Which of my past ads had the best cost per purchase? Use what worked there in the new angles.

Step 6: Write it up

Put the whole plan on one page: objective, audience, budget and structure, the three ad angles, what we'll measure, and when we'll check results.

Save that. It's your checklist when you build, and it's what you compare against in two weeks.

What to do next

• Build it: create a campaign from scratch. New campaigns start paused, so you can check everything first.
• Building for a client from a written brief? See campaign from a brief.
• Two weeks in, run a campaign performance audit.

Common mistakes

• Planning without the Pixel. If you want sales but the Pixel isn't recording purchases, Meta can't optimize for them. Fix that first.
• Too many ad sets for the budget. Each one needs enough spend to learn.
• Audiences that are too small. A tiny audience gets expensive fast and stops delivering.
• Treating estimates as promises. Audience sizes and any results the AI predicts are rough guides. Your real numbers will differ.

FAQ

Will planning change anything in my account? No. Every step reads data or asks Meta for estimates. Nothing is created until you turn on write access and ask.

Can AdChat tell me how many sales I'll get? It can give rough expectations based on your past results, if you have any. Nobody can promise sales numbers before a campaign runs.

I've never run ads. Is this still useful? Yes. Tell the AI that up front and it will explain each choice in plain English. A few one-time steps, like adding a payment method, happen in Meta itself, and the AI will tell you what to click.

Try it free for 7 days. Create your account and plan your next campaign before you spend a cent. Questions? Email [email protected].