How I Created an AI Agent That Posts to Facebook While I Sleep

You set it up once, walk away, and the next morning your Facebook page has a new post waiting for you. Text written by AI. Image generated by AI. Uploaded automatically. Zero manual work after the first 20 minutes.

That is not a fantasy. I built this exact system using three free or low-cost tools and five steps inside Make.com. The niche I used as a live demo was dog treats for beginners, and by the time the automation ran, a real post with a real DALL-E image was sitting on my Facebook page. This walkthrough shows you every step, the exact modules to add, the variable you almost certainly will miss, and the timing trick that keeps Facebook from flagging your account.

What You’ll Walk Out With

A complete Make.com scenario that writes a Facebook post, generates an image, and uploads both automatically
The exact OpenAI module settings for text generation and DALL-E 3 image generation
How to use the Set Variable module to convert OpenAI output into a string Facebook actually accepts
The HTTP “Get a File” step that downloads the image URL to the cloud without touching your computer
A safe posting interval strategy (13 or 17 hours) that avoids Facebook automation flags
Three real use cases you can plug this into: affiliate marketing, digital products, and mailing list growth
How to find which niche to use and whether this works for your audience (use finder.platformproof.com to match your skills to the right online income path)

Why Facebook in 2025

The common objection before anyone builds this system is: “Who even uses Facebook anymore?” Fair question. I looked it up on SimilarWeb and the answer is: 11.92 billion visits last month, up nearly 10% from the month before. That is not a declining platform. That is still one of the largest traffic sources on the internet, and the pet niche, the health niche, the money niche, the parenting niche, every high-intent buying audience you can think of, all of them have active Facebook pages and groups you can reach.

The goal of this automation is not to go viral. The goal is attention and traffic, consistently, without you manually creating content every day. Once you have traffic flowing to a Facebook page, you can redirect it toward affiliate links, a digital product, a mailing list signup, or a lead magnet. This system is the top of the funnel. What you do with the traffic is the business.

The Three Tools You Need

Before touching any of the five steps, you need three things set up. None of them require a credit card to start:

Make.com: The automation platform that connects everything. You can start on the free plan. I have an affiliate link in the YouTube description if you want to use it.
OpenAI API account: Not ChatGPT Plus. The API. Go to platform.openai.com, create an account, get your API key. Takes about 30 seconds. You pay per use, and for the volume this automation runs at, the cost is minimal.
Facebook Business Page: Not a personal profile. A dedicated business page for your niche. Easy to create and free.

With those three in place, the actual Make.com setup takes 15 to 20 minutes. Here is every step exactly as I built it.

Step 1: Create a New Make.com Scenario

Log into Make.com and click “Create a new scenario.” Give it a name. I named mine “Facebook Posting Page” so it is easy to find later. Do not skip naming it. When you have 20 automations running you will thank yourself for the organization.

Hit the plus sign in the scenario builder. This is where you add the first module.

Step 2: Add OpenAI – Create a Chat Completion

In the module search, type OpenAI. Select “Create a Chat Completion.” Connect your OpenAI API key. If you have not done this before, click Add, enter your API key from platform.openai.com, and save. It takes 30 seconds.

Once connected, select a model. Pick the latest available one in the dropdown. It does not matter exactly which model as long as you are not using something outdated. Then write your prompt in the Message field. Set the role to User. The prompt I used for dog treats was:

“Write a short, engaging Facebook post based on our niche: dog treats for beginners.”

Substitute your actual niche where “dog treats for beginners” appears. If you have not picked a niche yet, previous videos cover how to find low-competition niches using AI. The prompt is intentionally simple. ChatGPT will add emojis, hashtags, and a short hook on its own. You do not need to engineer it beyond this.

Click Save. You can run the module once at this point to confirm the output looks right, but it is not required before moving to step 3.

Step 3: Add OpenAI – Generate an Image With DALL-E 3

Click the plus sign again to add a second module. Search OpenAI again. This time scroll down and select “Generate an Image.” Choose DALL-E 3 as the model.

For the prompt field, you are going to pull in the output from the previous step. Click the field and select from the dropdown: Choices > Message > Content. This tells DALL-E to generate an image based on whatever the chat completion just wrote. The image will be thematically connected to the text post, which is exactly what you want for visual cohesion.

Click Save. If you run the scenario at this point, you will get a text post and an image URL as output. The image URL is a temporary hosted link from OpenAI. That link is what the next two steps use.

Step 4: Set Variable + HTTP Get a File (The Step Most People Miss)

Here is where a lot of people run into a wall. If you try to connect the OpenAI image output directly to a Facebook module and upload it, you will get an error. The output from OpenAI is not in a format Facebook understands. You need to do two things before the Facebook connection.

First: Add a Set Variable module. In Make.com, click Tools, then Set Variable. Drag it so it sits between the DALL-E module and your Facebook module. Give the variable a name, something like “chat_var.” For the value, map it to: Choices[0] > Message > Content from the chat completion output. This converts the array output into a plain string that downstream modules can use as caption text.

Second: Add an HTTP Get a File module. Click the plus sign, search HTTP, and select “Get a File.” For the URL field, map it to the image URL from the DALL-E output. Specifically: Data > URL. This step downloads the image from OpenAI’s temporary URL to Make.com’s cloud. Your computer never touches it. The file now exists in the scenario’s memory and can be passed directly to Facebook.

Click Save. The scenario now has four modules: Chat Completion, Set Variable, Generate Image, HTTP Get a File. One more to go.

Step 5: Connect to Facebook Pages and Create the Post

Add one final module. Search “Facebook Pages” and select “Create a Post with Photos.” Click Add to connect your Facebook account. Walk through the Facebook authorization flow. It takes about a minute and Make.com will list all the business pages tied to your account.

Select your target page from the dropdown. For the Photo field, select the output from the HTTP Get a File module. For the Caption field, map in the “chat_var” variable you set earlier. That is the clean text string from the chat completion. Optionally, add a comment like “Learn more” to drive engagement, reply to DMs, or direct people somewhere specific.

Click Save. Click “Run Once” to test the full flow. Here is what should happen: Make.com generates a text post, creates a matching DALL-E 3 image, downloads it, and uploads both to your Facebook page. In my live test with the dog treats niche, every module showed a green checkmark and the post appeared on the Facebook page within about 45 seconds.

Not sure which niche or business model fits your actual situation?

Answer a few quick questions at finder.platformproof.com and get a personalized match for your first $3,000 online based on the skills you already have.

Setting the Schedule (And the Timing Trick That Matters)

After the test run works, you need to set this scenario to run automatically. In Make.com, you control this with the scheduling settings in the scenario itself. Do not run this every hour. Facebook technically does not allow automation, and if you post too frequently in rapid succession, your page can get flagged or restricted.

The safe approach: set the interval to every 13 hours or every 17 hours. The odd numbers matter. Facebook’s detection systems look for mechanical regularity. Posting every 12 hours looks like a bot. Posting every 13 or 17 hours looks more like a human who posts twice a day with irregular timing.

Even better: add a randomized function to the scenario’s trigger timing. In Make.com you can write a simple expression that generates a random number within a range and only fires the scenario if that number hits a specific value or falls within a threshold. This adds natural variation to your posting timing that is nearly impossible for automated detection to flag. It is an optional step but worth building in from the start.

Honest Drawbacks

This is a genuinely useful automation, but there are real limits worth naming before you build your entire business on it.

Facebook’s automation stance is real. The platform technically prohibits third-party automation for posting. Make.com uses the official Facebook Graph API, which is allowed for business pages, but you are still operating in a gray zone. Post too frequently or behave too mechanically and you risk page restrictions.
DALL-E image quality varies. The images are good enough for engagement posts but rarely look like professional branded content. If your niche requires polished visuals, you may want to add a Canva API step or review images before they post.
OpenAI API costs money. Not much, but it is not free. At the volume this runs (one or two posts per day), costs stay well under a few dollars per month. Scale to dozens of posts per day and that changes.
This gets you traffic, not conversions. A Facebook post with a nice image gets likes and maybe follows. That traffic needs somewhere to go. Without a clear next step (affiliate link, opt-in page, product), the automation generates activity but not income.
Niche selection is still the hardest part. The automation works for any niche, but a post about a bad niche still generates zero business. The five steps above assume you have already chosen a viable niche. If you have not, that work comes first.

What to Connect This To

The automation is the traffic machine. The business sits on top of it. Here are the three setups that make this worth running:

Affiliate marketing: Include an affiliate link in the post caption or in a comment. Every post that drives a click earns a commission if someone buys. The dog treats niche, for example, has dozens of Amazon affiliate products and dedicated pet affiliate programs paying 8 to 15% per sale.

Digital products: Use the posts to drive traffic to a landing page for an ebook, a template, a mini-course, or a coaching offer. A Facebook page with consistent content builds trust faster than cold traffic. The automation runs the content calendar while you build the product.

Mailing list growth: Direct post traffic to a free lead magnet opt-in. A checklist, a short guide, a swipe file. Once someone is on your list, you own that contact. Facebook can change its algorithm or restrict your page tomorrow. Your email list is yours. Using this automation to grow a list rather than just a follower count is the more durable long-term play.

A Real Numbers Look at the Flow

Here is what the economics look like at a basic operating level, using realistic estimates rather than inflated projections:

Posts per day: 1-2 (at 13 or 17-hour intervals)
Make.com cost: $0 on free plan for low operation counts; $9/month if you scale up
OpenAI API cost: Roughly $0.01-0.03 per post (text + image generation); under $2/month at 2 posts/day
Setup time: 15-20 minutes the first time; near zero ongoing
Maintenance: Check in once every few days to make sure no errors have appeared in the scenario run history

The tool cost is not what determines whether this works. The question is whether your niche has enough of an audience on Facebook and whether you have a clear path to monetize the attention once you have it.

Find Your X

This automation works for any niche, but the niche choice is the variable that determines everything. The right niche for you depends on what you know, what you can speak to credibly, and what audiences are already buying on social media. If you are not sure which direction to take this, finder.platformproof.com will match you to the right starting point based on your specific situation. Answer a few questions and get a recommendation you can act on this week.

Frequently Asked Questions

Does this work with a personal Facebook profile or only a business page?

The Make.com Facebook integration works with Facebook Business Pages, not personal profiles. You need a dedicated page for your niche. Pages are free to create and do not require a business registration. Just go to Facebook, click Create, and choose Page.

Can I use GPT-4 instead of the default model?

Yes. In the OpenAI Create a Chat Completion module, the model dropdown lists everything your API account has access to. Select GPT-4 or GPT-4o if you want stronger output quality. The cost per post goes up slightly but stays well under a few cents per run.

What happens if the DALL-E image URL expires before the HTTP module downloads it?

OpenAI’s image URLs expire after about an hour. In a single Make.com scenario run that completes in under a minute, this is not a problem. The HTTP Get a File module downloads the image within seconds of it being generated. If you ever see an error saying the URL is invalid, check whether you have added any delays in the scenario that would cause a gap between generation and download.

How do I prevent Facebook from flagging my page for automation?

Post at irregular intervals rather than on a fixed schedule. The video recommends 13-hour or 17-hour gaps rather than a round 12 or 24 hours. Optionally add a randomized function in Make.com to vary the timing further. Also avoid posting more than twice a day, keep the content relevant to your page’s niche, and respond to real comments manually when they come in.

Do I need a paid Make.com plan?

Not to start. The free plan includes enough operations for one or two posts per day. If you scale to multiple pages or more frequent posting, the Core plan at $9/month provides significantly more operations and removes some free-tier limitations. Start free, upgrade when you have a reason to.

What if I want to post to a Facebook Group instead of a Page?

The Make.com Facebook Pages module is specifically for pages. For groups, you would need the Facebook module that handles group posts, or a workaround using the Facebook Graph API directly through Make.com’s HTTP module. Group automation is more restricted by Facebook than page automation, so tread carefully with that setup.

Can I use this for multiple niches at once?

Yes. Build a separate scenario for each niche and each Facebook page. Each scenario is independent. You can run a pet niche page, a fitness niche page, and a personal finance niche page all from the same Make.com account as long as you stay within your plan’s operation limits. Just name the scenarios clearly so you can track which is which.

What should I do after the automation is posting consistently?

Pick a monetization path and build the funnel that connects your posts to revenue. The most common options are adding affiliate links to post captions, linking to a lead magnet to grow an email list, or directing traffic to a digital product page. The automation handles the top-of-funnel content creation. Your job is to build the path that turns that traffic into income.

Sources

SimilarWeb traffic data: Facebook.com, 11.92 billion visits, up ~10% month over month (referenced in video)
Make.com, automation platform used in this walkthrough: make.com
OpenAI API, text and image generation: platform.openai.com
OpenAI DALL-E 3, image generation model used in step 3
Facebook Graph API, underlying API Make.com uses for business page posting

Helping 1 million working adults make their first $3,000 online with the skills they already have. Alston Godbolt, Platform Proof.