How AI Agents Make Affiliate Marketing EASY in 2025

Most people who want to start affiliate marketing get stuck at the same point. They spend hours on YouTube trying to figure out which programs are worth joining, then more hours trying to come up with content ideas, and by the time they have a plan they are too burned out to actually create anything. The research phase kills more affiliate marketing businesses before they start than anything else.

In this walkthrough, Alston builds a real AI agent inside Make.com that handles all three of those research jobs automatically. You give it a niche, it comes back with a ranked list of affiliate programs, ten content ideas sorted by buyer intent, and full content outlines dropped straight into Google Docs. The agent even has a name: John. And once you build it, you can trigger it again anytime you want fresh ideas, or you can sell the same setup as a service to other business owners on Fiverr.

What You’ll Walk Out With

A working Make.com scenario that finds affiliate programs in any niche, complete with commission rates, cookie durations, and affiliate-friendliness scores
An Airtable database that stores every program in its own row so you can sort and filter without spreadsheet headaches
A second agent step that generates ten bottom-of-funnel content ideas based on the programs your agent found
A third step that turns each idea into a full content outline saved to Google Docs, ready to film or write
A named AI agent called Johnny on the Spot that you can message on demand to run the whole process again
A clear bottom-of-funnel content strategy that puts your videos closer to the money from day one
A second business idea: selling this same agent as a service to other affiliate marketers on Fiverr
A way to figure out which niche is actually right for you before you build anything: finder.platformproof.com

Why Affiliate Marketing Research Kills Most People Before They Start

Affiliate marketing is not complicated in theory. You pick a niche, find products in that niche that pay a commission, create content that brings in people who are already looking to buy, and collect the commission when they click through and purchase. That is the whole model. But the research step, specifically finding programs that actually pay well, have reasonable cookie durations, and are open to new affiliates, can eat an entire weekend without producing a single piece of content.

On top of that, most beginners pick the wrong type of content. They create how-to videos and overview guides that attract people who are still figuring out what their problem even is. Those viewers are at the top of the funnel and they are far from ready to buy. The result is traffic that does not convert. Alston’s agent solves both problems at once: it finds programs worth promoting and then generates content ideas that are already close to the point of purchase.

Two Ways to Make Money With This Agent

Before getting into the build, it is worth understanding the two distinct ways you can use what you are about to build. The first is the obvious one: use the agent to run your own affiliate marketing business. You enter your niche, the agent does the research, you create content from the outlines it generates, and you earn commissions when your audience clicks through and buys. The agent removes the emotion from niche selection and program research so you stop second-guessing yourself and start publishing.

The second way is to sell the service. Imagine creating a gig on Fiverr where a client tells you their niche and you deliver a full affiliate marketing research package: ten vetted programs with commission rates and cookie durations, ten content ideas, and ten ready-to-film outlines. The agent does the work in minutes. You package the output and charge for your time setting it up and delivering it. Whether you are building your own income stream or adding a service to an existing business you already run, the same agent handles both.

The Tools You Need and What They Actually Cost

This entire system runs on tools that cost almost nothing to start. Make.com is free. It is the automation platform where you connect all the pieces together, and you can run dozens of scenarios on the free plan before you ever need to upgrade. The ChatGPT API through OpenAI costs less than five dollars to get started, and because each run of this agent only calls the API a few times, your costs stay very low. Airtable is free at the level you need here, and you are using it to store your affiliate program data and content ideas in organized tables. Google Drive and Google Docs are free if you have a Google account, which you almost certainly do. Cloud Convert comes into play later in the workflow when you want to convert file formats, but you do not need it for this initial build.

The total cost to run this system for the first time is under five dollars. The total time to build it from scratch, including all the setup steps Alston walks through in this video, is roughly thirty minutes. You do it once and then you have a repeatable research machine that you can run anytime you want.

Step 1: Build the Affiliate Program Finder in Make.com

Start by going to Make.com and creating a new scenario. Search for OpenAI in the app directory and select the “Create a Completion” module. Choose the latest available GPT model from the dropdown. The model name changes over time so just pick whatever is most current when you are building this. Add a message role set to “user” and paste in the affiliate program research prompt. The prompt tells the API to return ten affiliate programs in your niche, and for each one it includes the program name, commission rate, a direct link to apply, cookie duration, and an affiliate friendliness score.

Before you run the scenario, set up your Airtable table. Go to Airtable, open your base, add a new table called “affiliate programs,” and delete the default columns. Add these fields as single-line text: Affiliate Commission, Affiliate Website, Cookie Duration. Add a number field for Affiliate Friendliness. Add a date field for Date Added. Save the table. Now go back to Make.com and add the Airtable “Create a Record” module after your OpenAI step. Connect your Airtable account and select your base and the affiliate programs table.

Here is where most beginners run into their first problem. When you run the scenario without any processing step in the middle, the output from ChatGPT comes back as one large block of text. Airtable cannot read that cleanly and you end up with all ten programs crammed into a single row instead of ten separate rows. To fix this, add the built-in JSON “Parse JSON” module between your OpenAI step and your Airtable step. Set its parse string to pull from the result of the OpenAI completion. This cleans the output and makes it machine-readable. Now when you run the scenario, Airtable gets one record per program. Sprocket, Printiful, Zenrop, and others each get their own row with their commission rate and cookie duration filled in cleanly.

One practical tip from the video: remember to save your Make.com scenario early and often. Click the disc icon in the interface every few minutes. If you close the tab without saving you will lose your work, and rebuilding from scratch wastes the time you just spent getting the configuration right.

You can also customize the prompt parameters. Want only programs that pay at least a twenty-five percent commission? Add that to the prompt. Only interested in programs with at least a one hundred and twenty day cookie window? Add that too. The agent will filter its results based on whatever specifications you give it, which means you can tighten or loosen the criteria depending on your niche and your goals.

The Bottom-of-Funnel Content Strategy That Actually Converts

Before moving to Step 2, it is worth understanding the content strategy baked into the agent’s content idea prompt. Most beginner affiliate marketers start at the top of the funnel. They make how-to videos, complete guides, and educational content aimed at people who are still trying to understand what their problem is. That traffic is real but it is far from buying. The people watching a “what is dropshipping” video are not ready to purchase a dropshipping tool this week.

The content types that are closer to the money are review videos, unboxing videos, and first impressions. After that come comparison videos in the format of Product X versus Product Y. Then alternatives videos, where someone is already using a product and wondering if something else is better. Then best-of lists like “best dropshipping tool for beginners” or “best tool under fifty dollars.” These content types reach people who already know what they want and are just deciding where to buy it. Those are the people who click affiliate links.

When Alston ran his content idea prompt for the dropshipping niche, the agent generated ideas like “Is Zendrop Worth It in 2024?” and “Shopify Dropshipping Store Review: Is It Really Worth It for Beginners?” Both of those are review-style titles aimed at people who are already considering a specific product. That is the bottom of the funnel. That is where affiliate commissions happen.

Step 2: Generate Content Ideas Automatically

Add a second OpenAI “Create a Completion” module to your Make.com scenario. Use the latest model again. Add the content ideas prompt as a user message. This prompt is configured to generate ten YouTube content ideas in your niche, using the bottom-of-funnel framework described above. Once again, you will need a Parse JSON module between the OpenAI step and the output, because the raw API response will not map cleanly to Airtable without it. This time, you also need to set a data structure so Make.com knows how to separate the ideas into individual records.

Add a new table in Airtable called “content ideas” and keep it simple. One field for the title is enough to start. You can always add more fields later, like target keyword, video length, or thumbnail idea, but for this initial build the title is what you need. Connect the Airtable “Create a Record” module, point it at your content ideas table, and map the title field from your JSON output. Run the scenario and you will see ten content ideas appear in Airtable, one per row, each one ready to act on.

One thing to watch for: make sure your niche is correctly set in the prompt before each run. In the video, Alston accidentally had his prompt still referencing the pet niche when he switched to dropshipping, which gave him a batch of dog-related content ideas instead of dropshipping ideas. It is a quick fix but it is easy to miss if you are moving fast.

Step 3: Create Full Content Outlines Saved to Google Docs

This is where the agent becomes genuinely useful for content creation. Add a third OpenAI “Create a Completion” module. This time the prompt takes the content idea title from Step 2 and generates a full content outline for that video or blog post. The prompt produces a structured outline with sections, talking points, and an order that makes it easy to sit down and film or write without having to think about structure at all.

After the OpenAI module, add another Parse JSON step to clean the outline output. Then add a Google Docs module to create a new document. Connect your Google account, select the destination folder in your Google Drive, and map the title and outline content to the document. When you run the full scenario, Make.com calls OpenAI three times, once for programs, once for ideas, and once for each outline, and deposits the finished outlines directly into a Google Docs folder you specify.

In the video, after running the full scenario, Alston opened a Google Doc titled “Shopify Dropshipping Store Review: Is It Really Worth It for Beginners?” and found a complete outline inside it. Sections, sub-points, suggested talking points. He could take that outline, paste it into ChatGPT for a more fleshed-out version, or just use it as-is to record a video. The output that used to take hours of research and planning came out of the scenario in minutes.

Not sure which niche to run this agent for?

That is the question that stops most people before they ever open Make.com. Find the answer at finder.platformproof.com and get matched to the niche that fits the skills and time you already have.

Creating a Named Agent: Johnny on the Spot

Once your scenario is running, Make.com lets you create a named AI agent and connect it to your scenario. Alston named his agent Johnny on the Spot, with the scenario name Johnny Affiliate Marketing. The agent uses the GPT Turbo model and gets a system prompt that tells it what it is responsible for: finding affiliate programs, generating content ideas, and creating outlines on demand.

The practical benefit of creating a named agent is that you can trigger the whole research process with a message instead of logging into Make.com and clicking run. You could set up a WhatsApp integration and send a message that says “Give me new content ideas for the dropshipping niche in 2025” and Johnny on the Spot runs the scenario and deposits the results into Airtable and Google Docs automatically. This is how a single person can operate what feels like a research team: one message, one agent, fresh ideas ready in minutes.

Building Toward a Team of Specialist Agents

Right now, one agent handles everything: affiliate programs, content ideas, and outlines. That is fine for getting started. But Alston mentions the natural next step for anyone who wants to go deeper. Instead of one agent that does everything, you build a team of specialized agents. One agent is responsible only for finding affiliate programs. A separate agent handles content ideas, with sub-agents that specialize by platform: one for YouTube ideas, one for blog ideas, one that searches Reddit for trending questions, one that monitors Google News for timely angles. Each specialist is better at its job than a generalist would be. You start with one agent that does everything and build toward a team where each agent does one thing very well.

That is the same trajectory as building a real team inside a business. You start as the person who does everything, and as you grow you hire specialists. With AI agents, the cost of adding a specialist is the time it takes to build the new scenario, which at this point in your learning is about thirty minutes per agent.

Honest Drawbacks to Know Before You Start

This system is genuinely useful, but there are a few honest limitations worth knowing about before you build it. First, the affiliate program data the agent returns is based on what the language model knows, not a live database. Programs come and go, commission rates change, and some programs the agent lists may not be accepting new affiliates right now. Treat the output as a starting-point list to verify, not a finished directory to act on blindly. Always click through to the actual program page before you apply.

Second, the content ideas are good but not personalized to your audience. They are solid bottom-of-funnel angles, but you still need to watch what is actually getting traction in your niche on YouTube right now. Use the agent’s output as a research foundation and then cross-check against current search volume and trending topics before you commit to filming.

Third, the thirty minutes Alston estimates is accurate for someone who has worked in Make.com before. If this is your first automation scenario, plan for longer. The JSON parsing step in particular trips up most beginners on the first try. That is not a reason not to build it. It is just a realistic expectation so you are not frustrated when the first test run produces an error.

Find Your X

Before you build an agent to find affiliate programs in a niche, you need to know what niche you are actually building for. Most people skip that step and end up running research for a niche that does not match their interests, their existing knowledge, or the time they realistically have to create content. The Finder tool at finder.platformproof.com is designed to match you to the right starting point based on what you already know and what you actually want to build. Get that answer first. Then build the agent around the niche that fits.

Frequently Asked Questions

Do I need coding experience to build this Make.com agent?

No. Make.com is a visual automation platform. You connect modules by dragging and dropping, fill in configuration fields, and paste in prompts. The hardest technical step is understanding the Parse JSON module, and that is fully visual once you see how it works. Alston builds the entire thing on screen in the video without writing a single line of code.

How accurate is the affiliate program data the agent returns?

The agent returns programs based on what the language model knows about your niche. It is a solid starting point, but commission rates, cookie durations, and program availability change over time. Treat the list as a research shortlist and verify each program directly on its affiliate page before you apply or start creating content around it.

Can I run this agent for multiple niches?

Yes. The niche is just a variable in the prompt. You can change it to any niche you want before running the scenario: dropshipping, personal finance, home fitness, language learning, anything. You can either update the prompt each time or, as Alston mentions in the video, you can pull the niche from a field in Airtable automatically so you do not have to hardcode it each run.

What does it cost to run this each week?

The ChatGPT API costs scale with the number of tokens you use. A typical run of this scenario, which calls the API three times to generate programs, ideas, and outlines, costs pennies per run. Running it once a week for a year would cost well under ten dollars in API fees. Make.com’s free plan covers most use cases at this volume. Airtable and Google Docs are free at this scale.

How is this different from just asking ChatGPT directly?

When you ask ChatGPT directly, you copy and paste results manually, format them yourself, and store them wherever you can find space. This Make.com agent takes the same API call and automatically stores the output in structured Airtable tables and Google Docs, formats it correctly, runs it on a schedule if you want, and can be triggered by a WhatsApp message. The output is the same. The difference is that everything is organized and reusable without any manual work on your end after the initial build.

Can I sell this as a service to other people?

That is specifically one of the two use cases Alston covers in this video. You build the agent once and then offer the research package as a Fiverr gig. The client gives you a niche, you run the agent, and you deliver a formatted report with vetted affiliate programs, content ideas, and outlines. You are charging for your time setting it up and delivering the output, not the agent’s time. As you refine your prompts and your workflow, the quality goes up and the time it takes you goes down.

What comes next after building this agent?

Alston mentions the next planned additions to this workflow: a lead magnet generator, email marketing automation, and an analytics agent that tracks what content is performing. The idea is that each new agent handles one more piece of the affiliate marketing system, so over time you have an automated backend for the entire operation, not just the research phase.

What type of content should I start with in a new affiliate niche?

Start at the bottom of the funnel. That means review videos, unboxing videos, and first impressions content about specific products. Then move to comparison videos like “Product A versus Product B” and alternatives videos for people who are already using something and looking for a switch. These content types reach people who are close to buying, which is where affiliate commissions actually come from. Save the broad how-to and educational content for after you have some traction and an audience that trusts your recommendations.

Sources

Alston Godbolt, “How AI Agents Make Affiliate Marketing EASY in 2025,” YouTube, Platform Proof channel
Make.com: automation platform used to build all agent scenarios in this walkthrough
OpenAI API: GPT-4o used as the AI reasoning layer inside Make.com
Airtable: free database used to store affiliate program data and content ideas
Google Docs: destination for generated content outlines

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