Its been six months since I first started working with Outlier AI. I originally found the platform through Indeed, and applied not expecting much. There have been some highs and lows, but as Outlier AI has become one of the most popular companies on GigFish, I feel compelled to write about my experience with them in more detail.
Onboarding & Initial Assessment
It began with a simple application process. I filled out the application form, uploaded my resume, and waited for a response. A few days later, I received an email inviting me to complete a short assessment. It was designed to essentially test my ability to follow instructions, and answer multiple choice questions about whether an AI response fulfilled a given prompt based on several categories.
I didn't really take the assessment too seriously, and a few days later I was notified that I had failed, but I was given the opportunity to retake it. I did not opt to retake it, and continued my job search elsewhere. Surprisingly, a month later, I received an email inviting me to onboard to the platform as if I had passed the assessment. This would be my first hint of just how disorganized the platform could be.
The onboarding process was much the same as the previous assessment. It involved pages of information about the project guidelines, rubrics regarding how to score the responses, and multiple choice questions based on the guidelines. It concluded with a couple of practice tasks with feedback about what was marked incorrectly if any of the answers were incorrect.
Starting the First Project
I received my first project assignment immediately after onboarding. The tasks involved grading AI responses from a chatbot designed to help people with their travel plans. I would read through a prompt and two AI responses, then grade the responses based on their ability to follow instructions, provide truthful information, verbosity, writing quality, and overall score.
At the end, I would compare the two by providing a "likert" score on a scale of 1 to 7. "1" being response A was much better, "7" being response B was much better, and "3" if they were about the same. Finally, I would justify my score with a paragraph about which was better and why.
Each task paid $25/hr and I was allowed a 60 minute window to complete the task. About one month into the job, the pay rate was increased to $30/hr. The work was steady, predictable, and I was given the opportunity to work towards bonuses that were based on the number of tasks completed.
At this point, I was working full time on the platform and could make anywhere from $1,500 to $2,000 per week. On weeks where bonuses were available, I could make upwards of $3,000. Things were going well, and I was grateful.
Project Hopping & Disorganization
After a few months having a solid, dependable stream of work, the platform started to change for the worse. I would onboard to a new project, take the assessment, and work for a week or two. Then, the project would either abruptly end or I would be removed following a string of questionable reviews from other contractors.
It was not such a big deal as more work was still available and new projects generally replaced old ones immediately. Around this time, the "marketplace" feature was introduced to the platform, where you could theoretically pick from different projects and see the task time limits and pay rate in advance.
More often than not, the projects listed there would be unavailable, full, or have no tasks available. This was frustrating, as the marketplace mostly seemed to give the illusion of choice, when in reality, nothing was actually available.
Throughout this time, I was also experiencing a lot of disorganization on the platform. I would receive emails meant for other contractors, and at times they would mistakenly say I had been removed from a project or assigned to another. During the summer of 2024, these occurred frequently. So frequently, that I became conditioned to ignore them and would login to my dashboard to confirm what my actual status was.
Despite all this, I was still making a good amount of money, and was never without work for more than a day or two. However, as the months went on, projects became shorter, tasks were given less time, and bonuses were less frequent. Moving forward, I would never make as much money as I had in those late summer months.
The Referral Program
From the beginning, Outlier AI has had a referral program. You could refer friends and family to the platform by providing their email address, and if they completed 10 hours of work, you would both receive a bonus. Beginning in the fall of 2024, the referral program introduced links that users could share to more directly refer people to the platform.
The referral program boasts an incredible potential payout of up to $500 per referral. However, the reality is that it is extremely difficult to secure payouts. There have been many instances where users would refer people, see a pending payout of $100-$500, and then have the payout mysteriously decrease to a measly $5 without any warning. This is well documented and can be witnessed as a common complaint on the Outlier AI subreddit.
At the time of writing, GigFish has referred over 300 people to the platform. Of those 300, only 4 people been given tasks, and only 1 person has been given a payout. To put this into perspective, only 1.3% of referrals have been given tasks, and only 0.3% of referrals have been given a payout. This is also a common complaint that can be seen on the Outlier AI subreddit.
The Final Verdict
There's no denying that Outlier AI is a legitimate platform that pays you to train AI models. However, it is not a stable gig, work is not always available, and it becomes increasingly difficult to make a consistent income with each passing month. The platform is disorganized, the support is lacking, and the referral program is borderline fraudulent. While myself and those on the GigFish staff have had success with the platform, we cannot in good conscience recommend it as a stable gig to pursue in 2025.
You may be one of the lucky ones who can make a consistent income with the platform, but if things do not improve, I would recommend trying your luck elsewhere. There are many other platforms that pay you to train AI models, and many of them are becoming more reliable. However, when there are tasks available, Outlier AI is one of the best paying platforms out there. If you've got the time, it may be worth your while to check out their job board and give it a shot. Just don't expect it to be a stable source of income.
