Welcome back to Champions of Performance Marketing. This is Part 2 of my conversation with Luke King, founder of affLIFT.
If you haven’t read Part 1 yet, I’d recommend starting there.
We covered how affLIFT was built from a side project into a full-time business; how Luke personally reaches out to every member who cancels; which traffic sources and verticals are working right now; and why affiliate marketing has never been dying, just constantly moving.
In this second part, we get into the affiliate campaign optimization. Landing pages or direct links, and when to use which.
How many creatives to test on push traffic, and what the budget math looks like when you add more variables.
What ROI number does Luke actually use to decide whether a campaign is worth scaling? And then we go deep on the AI assistant he’s built, a system running on Claude that monitors campaigns, creates whitelists, and keeps an eye on forum sentiment while Luke is away from his desk.
We also talk about what he thinks is the biggest mistake someone can make when starting in affiliate marketing. It has nothing to do with traffic sources or creatives.
Listen to the episode on our Spotify channel
Q: When you run an offer, do you start with a direct link or do you use a landing page?
Luke King: It really depends on the offer.
Let’s use dating as an example, since it’s a vertical I have experience with. If it’s a dating offer with a direct registration page, which isn’t as common anymore but used to be very popular, I might create a pre-lander just to warm up the lead before they get to the registration page.
Take a look at what you’re promoting. That’s rule number one.
After running enough offers, you’ll start to get a feel for it. This landing page looks similar to one I ran a year ago that performed well, so I’m going to go direct. Or this one looks like one that didn’t perform well, so let me try something in front of it.
A lot of it comes from experience. For someone who isn’t sure, my recommendation is almost always: yes, test a landing page. Running a landing page split tested against a direct link will tell you pretty quickly which one performs better. If the direct link wins, remove the landing page and go direct.
If the landing page wins, maybe split-test a few more variations. You can use a tracker like CPV Lab to manage all of that easily.
The general recommendation from me and the community leaders on affLIFT is: it can’t hurt. If you can spin up a landing page and test it, you should, because seven times out of ten, the landing page makes a difference.
Q: So you wouldn’t recommend just testing the direct link first to see if the offer converts, then building a landing page?
Luke King: That’s a fair point.
If the offer isn’t converting at all, it doesn’t really matter what you put in front of it.
A landing page isn’t going to fix a broken offer.
But at the same time, if you run 50 percent to a landing page and 50 percent direct, there’s always the possibility that the direct link won’t perform without something warming the visitor up first. If you only ever tested the direct link and it failed, you might write off the offer entirely, when in reality, it just needed a better pre-sale.
The motto in affiliate marketing is test, test, and test again. We can sit here and discuss theory all day, but the only real way to know is to actually run it.
Q: What about creatives? For push traffic specifically, how many should someone test for a campaign?
Luke King: It depends, but with push, I typically start with three to five creatives with different looks and different combinations of image, title, and description.
With AI, it’s so much easier than it used to be. You create one creative idea, drop it into an AI tool, ask for four more variations, and within ten seconds, you have what you need.
Back in my day, you had to do all of that manually, which was a real time constraint. Now it’s almost instant. More is generally better, and three to five with push creatives specifically is what I do pretty much every single time.
With landing page creatives, AI also makes the process easier, but I’d say be careful.
A lot of AI-generated landing pages still look like they were built by AI, depending on which model you use. When that happens, the results tend to be disappointing.
Sometimes AI produces something great. Sometimes it doesn’t.
And sometimes the ones you least expect to perform actually do.
Which brings it back to this: you have to test.
For landing pages, I’ll do at least one to run the 50-50 split test, and adding a couple of extra variations is worth it when your budget allows. My personal AI assistant, for campaigns where I’m not directly linking, will do three to five landing page variations by default.
If your daily budget is $20 and you’re split testing two landing pages, each one gets $10 worth of traffic. Add five landing pages and that drops to $4 each. The amount of data you get for $4 is a lot less than what you get for $10. This is something that intermediate and beginner affiliates regularly get wrong. They’ll run $2 of traffic and say the offer isn’t converting. You don’t have enough data to say that.
The more variables you add, the more budget you need. But when you do find something that works, you have everything in place to scale it.
Q: How do you decide which campaign to scale? What metrics are you tracking?
Luke King: ROI is number one. Profit correlates with that.
But when I’m running a landing page, I’m also watching click-through rate and EPC, meaning how much I am earning every time someone clicks through from the landing page.
EPV too: how much am I earning per landing page visit overall?
There are a lot of different ways to measure a campaign in affiliate marketing, and a lot of different ways to define what success looks like.
Is a 10 percent ROI on a campaign generating $100 a day worth it? It depends on the affiliate and their profit goal.
And people forget that there are real costs attached to every affiliate campaign that have to be covered before that profit number means anything. Your tracker, your affLIFT membership, all of that adds up.
30% ROI is kind of my cutoff. I’m looking for 30%. If I’m below 30%, it’s possible that I’m going to stick with it; it might still be worth running, but typically, if I’m optimizing a campaign and trying to scale, I want to see at least 30% ROI.
Early in a campaign, if I’m using a landing page, I’m not even focused on conversions yet.
I’m just trying to get my click-through rate up and figure out whether I’m getting enough data to form a real opinion. Once I see that I am, then I start looking at conversions, working toward positive ROI, trying to reach that 30 percent target, and then seeing how far I can scale from there.
Q: You mentioned you have a personal AI assistant you built. What does it do?
Luke King: You can call him “he.” His name is Jeff.
I started experimenting with AI a couple of years ago. We created Einstein, who became the AI community leader on affLIFT.
Early on, I decided that if you’re going to have an AI on a forum, you might as well give it a persona, a name, something people can reference easily.
About a few months ago, I started building something more complex. Jeff runs on Claude, specifically Opus 4.6 from Anthropic, and he lives on Telegram. I talk to him throughout the day across different projects: affLIFT, side projects, and affiliate campaigns.
I haven’t written the code myself. Jeff has done the programming, but I walked him through exactly how I monitor affiliate campaigns and what I’m looking for.
He goes through my stats on Propeller Ads and flags campaigns below 30% ROI that are still profitable. If something is well above 30 percent, he flags that too. Those are the rockets.
If something is performing poorly and I haven’t already paused it, he’ll pause it or send me a notification.
Years ago, Charles Ngo, a well-known super affiliate, wrote an article on his blog about creating systems for your workflow. I implemented a lot of what he shared in how I run my businesses. Now I’m trying to hand those same systems to the AI.
The beautiful thing about AI is that it will do it every single day, no matter what. I can be sick in bed or at the beach and Jeff still runs through those workflows. All I have to do is check in.
Jeff also monitors the forum throughout the day. If someone replies to a post I made and the response reads as negative or frustrated, Jeff sends me a Telegram message flagging the sentiment, so it gets my attention instead of sitting in an email inbox I might not check for hours. I’m adding to his responsibilities slowly and deliberately. He’s only been monitoring the forum for less than a month. The idea isn’t that AI needs to do all of your tasks for you.
The idea is for it to be an assistant that improves the tasks you’re already doing and makes them easier to manage. I think everybody’s probably going to have a personalized AI assistant like this eventually.
Q: You could have Jeff handle the cancellation outreach, too. Why haven’t you automated that?
Luke King: I could, and it would be easy to set up. But I don’t think it’s necessary, and the reason is that I want that to remain actual human contact.
That outreach is partly about genuinely trying to help, but it’s also about understanding why someone left. If an AI is doing that for me, I lose the human element of it.
This connects to something I did just last week. I updated the affLIFT terms and rules for the first time in eight years. One of the main reasons was to add a section specifically about AI. The forum is good because of its members sharing their actual experiences.
If all of a sudden the only content you see on the forum is people just regurgitating what AI tells them, then we kind of lose what makes it worth anything. There’s no more human experience. It’s all just AI.
Q: Are you actually seeing AI agents show up in communities like affLIFT?
Luke King: We’ve seen it come into the forum, yes. People messing around with agents, testing what happens when they send them into a community. But for the most part, our members are calling it out themselves. The response from the community is pretty clear: we don’t want this; it’s not the purpose of the forum.
Paying to get your AI agent in the forum is useless. If all of a sudden the only content you see on the forum is people just regurgitating what AI tells them, we lose what makes it worth anything.
What is kind of cool, though, is that we’re getting a lot of traffic from ChatGPT, Claude, and other AI platforms that are training off of publicly available forum data.
Most of affLIFT is behind a paid membership, so most of it isn’t accessible to those models. But they do have access to some of it.
And we’re starting to see referrals come through, meaning the AI is using what it has from affLIFT to help people who are typing affiliate marketing questions into these tools. It’s kind of cool that the AI models trust the information from the forum enough to use it to help people. Hopefully those people will come to the forum, sign up, and see what it’s really about, and then come learn more from the resources that are for paid members.
Q: What should affiliates automate first when they start using AI in their campaigns?
Luke King: One thing that’s working really well for me right now, one I’ve never really enjoyed doing manually, is creating multiple campaigns for a single offer. Here’s how it works in practice.
Say there’s a survey offer that accepts traffic from every country in the world, and I want to test ten different countries.
I go to the traffic source and look at their traffic data: impressions, average CPM, and anything available. I copy and paste that into a spreadsheet and hand it to the AI.
Then I take the offer data. Many affiliate networks show you the average eCPM, which is what affiliates are earning per thousand impressions, and put that in the same spreadsheet.
Then I tell the AI: here’s the available traffic; here’s the offer data. Which ten countries should we test, and go create those campaigns for me?
Traffic sources like PropellerAds now have APIs, so you can create campaigns programmatically. You give the AI your API key. Be careful with that, because you’re giving it access to all of your account data. But once you’ve done that, it can go and create your campaigns for you. What used to take me half an hour now takes less than a minute.
Train it as you would an employee. Have it create one campaign first. Check that it’s right. Then have it create the other nine. And have it document exactly what it did, so every time after that it follows the same process without you having to explain it again. Make sure you put in the correct targeting. If you mess up the bid, it’s going to cost a fortune, so double-check that. After the second time, you barely need to check.
Campaign creation is number one to automate. Optimization is the second obvious one. I’ve had AI-assisted optimization running on my campaigns for a while now, and it works well. AI is very good at patterns, not always great at raw numbers, but great at patterns. Tell it exactly what you’re looking for, have it go through the data, and it typically finds it.
Q: Can you give a specific example of what AI-assisted optimization looks like?
Luke King: Sure. In your tracker, it’s easy to pull up zones that are above 30 percent ROI. Just click a few buttons, and you can see them all. You can have the AI do that same thing. Give it access to your campaign data, and it goes through, finds every zone above 30 percent ROI, saves that list, creates a brand new campaign on PropellerAds targeting only those zones as a whitelist, and launches it. That’s something I’ve been doing manually for ten or twenty years. The AI does it in a few seconds.
And it can run that process all night. Every hour, every half hour, every fifteen minutes, it checks for new zones that are converting well and automatically adds them to the whitelist. Traffic sources are starting to build some of this into their own platforms, but having an agent that can do it across any traffic source is still a real advantage.
Q: Once you have a whitelist campaign running, do you pause the original?
Luke King: It depends on the traffic source and the campaign.
If you take those zones out of the original campaign and it goes negative without them, then yes, you’d probably just pause it. But you don’t necessarily have to remove them. Sometimes you leave them in at a lower bid and let the traffic keep running. If it stays profitable, you’re getting a bigger proportion of profitable traffic anyway, so it doesn’t hurt your bottom line.
There’s one thing I always tell affiliates to be careful of. If you set up a whitelist and only run those zones, you’re cutting off any new zones that might pop up and be profitable. Let’s say your whitelist has five zones. Any new zone that comes up, or any zone that was negative before but didn’t have enough data and might have turned profitable, and you’re no longer getting traffic from those.
On the forum, you’ll actually see members recommend blacklists more often than whitelists for that reason. If your goal is maximum ROI and you want tight control, run a whitelist. If you’re happy with a lower ROI and want more volume, run a blacklist: just remove the zones that don’t perform and let everything else continue. It just depends on your strategy and what you’re after.
Q: How important is tracking for affiliates today?
Luke King: Tracking is critical.
The whole point of running an affiliate marketing campaign is to end up with something profitable. Unless you’re lucky. Sometimes you put a campaign together, and it’s profitable without ever touching it. That’s the golden goose. But it typically doesn’t last long, maybe once a year if you’re running a high volume of campaigns, and even then, it doesn’t last.
The way you get there is by buying the data and then segmenting out what’s not working. That’s it. That’s the whole thing. You add different landing pages and different offers, but the end goal is a positive ROI, and the only way you reach it is by having the data that tells you what’s worth keeping and what isn’t.
That’s why on our beginner course, the number one thing we focus on is getting your first campaign live and learning how to track and optimize. That’s it. That’s affiliate marketing in its simplest form. You need the offers, you need the tools, but you only start making money when you’re reading the data and acting on it.
Q: Can AI eventually run a campaign entirely on its own, with no human involved?
Luke King: I think we’re still a long way from someone with no background in affiliate marketing being able to just create an agent, point it at a campaign, and have it produce profit on its own. I don’t see that happening anytime soon. You have to be able to control it, telling it what you’re looking for and why. The AI will know the basic structure of a campaign from what it’s been trained on, but it won’t necessarily know what to do in your specific situation.
And as we’ve already covered, this industry constantly changes. The AI will keep up to a point, but it’s not going to be as effective as having someone who actually knows what they’re doing behind the wheel. Someone who understands what’s happening has to be there to guide it. At least for now.
Once you can just tell the AI to go find a profitable campaign and run it entirely on its own, nobody will have an edge anymore. Because everyone will be able to do it.
Q: What are the rules around AI use on affLIFT?
Luke King: The forum is good because of its members sharing their real experiences. If that gets replaced by people posting whatever AI generates for them, we lose what makes it worth anything. My goal for affLIFT is to keep the human element at the center.
I can go to ChatGPT and ask it how to run a campaign in India, and it’ll try its best. But I’d much rather ask someone who has actually run campaigns in India and get their real answer. That’s what I think the main value of affLIFT will be going forward: it’s an actual human discussion forum where you’re talking to other people and getting experience that comes from having done the thing, not from having asked a machine about it.
You can use AI to help translate what you’re trying to say or to help structure it, but it needs to be your thoughts going in. You write your ideas; the AI helps you organize them. What’s not allowed is you type ‘create a guide on how to run pop ads in India’ into ChatGPT, it generates something, and you post that on affLIFT. There’s no value in that. Nobody learns anything from it.
Q: What’s your top advice for someone starting in affiliate marketing right now?
Luke King: The first thing I always say, and this has been my position since my days as an affiliate manager, is that affiliate marketing is hard.
There’s less misinformation around now than there used to be, but for a long time, people were sold on the idea that it’s easy, that you just press a button and you’re making $100 a day. That’s not the case. It’s never really been the case. Just like anything else, if you’re learning something new, there’s going to be a learning curve. Affiliate marketing is no different. It’s not a get-rich-quick thing. You have to set your expectations.
You need money to invest. You can’t really start from zero. And you need to accept that your first campaign probably won’t be profitable. Your second probably won’t be either. You might not be profitable until your tenth campaign, and it could take longer than that. If there’s something where I can pay $100 and instantly see $1,000 back, tell me, so I can do that. It doesn’t exist. What affiliate marketing is is an investment in learning.
Learn the trackers. Learn the tools. Get a real understanding of what actually goes into running a campaign. You can’t skip that foundation.
A lot of people come in and say, “I’m going to go run this on Facebook.” Great. But how are you going to do it?
We know what your end goal is: to be profitable. But what’s the plan for getting there? Having some sort of strategy, working it day to day, documenting what you’re doing so you can look back and know what didn’t work, knowing that you’re going to have to make constant little improvements. That’s the mindset.
If you can’t put together $100 to $200 to try this, wait. Don’t do it yet. Otherwise, you spend $25, get nothing, and walk away. You’ll never come back. Give yourself a real budget and a real shot.
And learn from other people. That’s one of the most underrated things you can do.
On affLIFT, our beginner course runs on about $10 of traffic. From being brand new to seeing your first conversions, $10 is enough to understand how it works if you’re running the right offer. We’ve had people join the forum and post in their intro thread that they’ve already spent $5,000 with nothing to show for it. That is not the goal. A fraction of that, spent learning from what other people have already figured out, goes so much further. Decrease the learning curve as much as you can just by looking at what other people are doing.
Where can people find Luke?
Connect with Luke King:
affLIFT.com (A-F-F-L-I-F-T.com) or at luke @afflift.com.
If someone has a question about affiliate marketing, AI, or anything related. I’m happy to respond. I think whatever success I’ve had with affLIFT and in this industry generally comes down to just wanting to genuinely help people. I’m still learning, too. I like having open conversations because that’s where everyone learns, including me. It doesn’t matter if you’re a forum member or a complete stranger. Just send me a message.
Conclusion
After this talk with Luke, a few things stayed with me long after we stopped recording.
One is how consistent his thinking is across every topic.
Whether he’s talking about why he emails every cancellation, why he built affLIFT at $20 instead of $100, why he won’t let Jeff automate the human outreach even though he easily could, or why he updated the forum rules to protect real discussion from AI-generated noise. It all comes from the same place.
He genuinely wants the community to be useful. Not impressive. Not profitable at the expense of usefulness. Just useful.
The other thing is how practical his AI thinking is. He’s not using it to replace judgment. He’s using it to execute the systems he’s already spent years building and refining.
Jeff runs the patterns Luke already knows to look for. Jeff creates campaigns the way Luke would create them. That distinction, with AI doing the execution of proven workflows rather than replacing the thinking behind them, is probably the most honest framing of where AI actually fits in affiliate marketing right now.
If you’re running campaigns and want to see how a 20-year veteran thinks through all of this in real time, affLIFT is where that conversation is happening.
CPV Lab ad tracker gives you real-time visibility, so you know what’s actually happening. Not what Google says is happening.
Start your 14-day free trial here!
Stay tuned for the next episode of Champions of Performance Marketing.
See the whole interview on YouTube or on Spotify
Author: Julia Draghici
Julia is the CEO of CPV Lab and CPV One ad trackers. She has 15+ years experience in the software industry, from development to management. For more than 7 years she is helping marketers get the best out of their marketing campaigns by using a performant ad tracker. Passionate about entrepreneurship, business and performance marketing, Julia loves helping people!
