Why We Reject Affiliates & Applications That Use AI

We reject most affiliate and influencer applications in the programs we manage that use AI to:

• Create content within their website
• Respond to our emails
• Auto respond to communications
• Answer user questions on their websites
• Apply to programs
• Generate images that involve branding

The main reason is that our clients expect us to manage their brand, what is said about them, and how the products or services are presented in the most accurate way possible.  It is impossible to catch and monitor everything, and AI makes it infinitely more difficult.  AI cannot quality control like a person, and AI does not fact check.  Yes there are always small exceptions to the rule like someone who is non-English speaking first but has English writers on staff, but they are few and far between.

Opinions on quality do not matter about how good the communication, content, or application sounds, accuracy is more important.

• LLMs are trained to give you the answers you want to hear, not what is required by the brand or factual.
  • Here’s an interesting study done on this.
• The datasets they have access to are likely not up-to-date so your content won’t be either.
• Even when an LLM or AI crawls a page, it formulates an opinion vs. giving factual information based on what it thinks you want to hear.
• These partners join a ton of programs and produce content in mass, there is zero way for them to quality control this.
• AI cannot handle advertising disclosures even though it pretends to, and this is something that comes up a lot.

Here’s a bit more information on each of them so you can decide if this is the same approach that you’d like to implement for your affiliate management, influencer, and ambassador processes.

Create Content Within Their Website

Multiple AI affiliates are applying to programs where they have LLM created content on their sites already.  In some cases they’ve already produced blog posts and YouTube videos including brand reviews, comparisons, and non-branded content that would normally be top-funnel.  Sounds great right?  Wrong.

Go to the content and look at what the content says.  This week I had to correct someone about product specs as they were not accurate.  The LLM decided to change the numbers, even though it “crawled” the page, and the LLM changed the safety of the product as it would not be used properly if the instructions were followed.  If the product was a child’s car seat and the LLM decided to replace the actual weight and size numbers, a child could get hurt.

The actual product was not for car seats, but it was about weight limits, and yes, people could get hurt.  Luckily we hadn’t approved that partner in, we did look at their site to find reviews and listcicles and that is where we discovered it.  It is our job to find this and try to correct it when we do.

The issue is these partners continue to churn out content that isn’t fact checked, or they have AI fact check and it gets worse.  Even if the content is factual when originally produced, I’m seeing a trend that they begin trusting the AI to edit and create more content, and my guess is because the AI has learned what they want to hear and now they automatically trust it is factual and the partner gets lazy.

Respond to Our Emails

Multiple partners respond to our emails with AI rewriting the content.  When they use AI to respond, its a telltale sign they’re now using it to create the content on their websites and social media too.

A few say it makes their words sound better or it saves them time.  While I appreciate the time saving aspect, the sounding better is not something we can accept.  This has lead multiple partners down the path of letting AI write the content, and then we run into the issues above.

Being an affiliate for a brand is a business relationship and the affiliate is responsible for knowing the brand, the product, and the service.  If they cannot send an email in their own words, how can they write about or include the product accurately in them?  If their words in an email are not clear enough, their written words on the page won’t be either.  An email is friendly and private, the content they produce will impact peoples’ lives including the customers and the brands.

Being an affiliate means treating the brands you partner with like it is your own company.  While the affiliate is not an employee, they are being trusted to be fair, accurate, and to take the time to properly promote the brands.  If the partner cannot take 2 minutes to write an email, they cannot be trusted to take the two minutes to fact check the promotion.  This is something that happens over and over, so email responses became something we remove or decline partners for.

Auto Respond to Communications

When there are errors in the copy or promotion or we need to get things updated, we need to talk to the affiliate or a person.  Some affiliates are using AI to auto respond to these emails, and those same partners don’t make the changes.  There has to be an open line of communication with the affiliate and the affiliate manager.

If this communication is blocked because AI is reading and responding to the emails, that communication is not happening and the changes that need made don’t get done.  This results in the affiliate manager getting in trouble with their boss or needing to take action on the partner.  The action here is turning off commissions or removing the partner from the program.  This is a business relationship and must be treated that way.

A huge pet peeve of mine is trying to get customer support when something is wrong with a product or service only to get stuck in a chatbot or AI loop.  This is what the auto responders for your email are doing.  The company and brand is better off removing the affiliate as the partnership has no way to correct mistakes and errors, and that could lead to larger problems if the relationship continues.

Answer User Questions on Their Websites

We’ve had a few affiliates build GPTs and chatbots to answer questions about the brand.  While they start out good, the information skews fast.  As products or services update and get new features or have different talking points, the chatbots don’t always.  This leads to them giving incorrect information and hurts the customer experience for the brand.

If the chatbot is using machine learning and AI systems that learn to power responses, this information can be skewed and become inaccurate quickly.  While the idea is a cool one, none of these systems are advanced enough to be accurate, so it is best left to the brand to answer questions while the affiliate does their job by pre-selling the product or service and sending the customer over.

One example that can have a serious impact is a food company with an affiliate program.  Maybe they have a new version of a product that now includes nuts.  Nut allergies can be deadly, and if the AI chatbot is saying they are nut free because the dataset is from 6 months ago and the affiliate isn’t checking for formula updates, people can buy it and get hurt.

I’ve seen a few companies each year that change their core offering like this vegan restaurant that began serving meat, or companies that change safety features and no longer support certain heights, weights, etc…  GPTs won’t update without the affiliate adding new data, and these affiliates are in hundreds of programs.  It is highly unlikely they’re checking every product, service, and brand they work with, and this is why we decline the sites using them.

Apply to Programs

This is a business relationship and the application should be the same as applying for a job.  If the affiliate is not going to take this serious, they’re not going to get approved.  Successful affiliates are not joining hundreds of programs, they join relevant ones.  There is no need for AI to fill out applications and write descriptions about how they’ll promote the brand.

If the partner cannot take 2 minutes to fill out a few sentences about how and why their website, app, social media and YouTube channel, or skool courses are a fit, how can the brand trust them to build quality content and promotions around the brand once they’re approved.

It is the same as non-responsive affiliate managers.  If they don’t respond when you want to promote the program and send them customers, why would they respond if something bad happens like untracked sales or the program going offline.

Generate Images That Involve Branding

Image creation is a tricky one.  As long as the AI is not taking someone else’s IP and using it, and everything is 100% original, it may be ok to use.  Only a licensed attorney can say for sure, and I am not one of those, so please talk to one.  What I do know is that I’m seeing AI distort brands.

The color of the logo, the sizes of the elements, font families, and other small mistakes that may not seem important to the affiliate, but they matter to the company’s legal and branding team.  If you’re using AI to create branded images, make sure it keeps the brand 100% in tact and make sure to get the affiliate manager’s sign off before pushing it live.

If I were going to do this, I’d ask the affiliate manager for the branding kit and I’d upload it and ask the LLM or AI system to follow the brand guidelines when using the logo, name, etc…  When we see applications and the brands have been distorted, that’s a big sign of a lack of quality control, and we reject the applications.

AI has a lot of good uses in affiliate marketing, but AI being used to create content, communicate with affiliate managers or for applications, and other tasks that require human review is not where it needs to be.  That’s why we reject affiliates using AI for specific tasks.