AI Training Guide · Step 4 of 5
AI Training Tips: How to Do the Work Well
AI training work is more than rating AI answers. Mercor describes it as reviewing AI answers in your area, creating examples that teach models how experts think, and checking quality against clear standards. Lionbridge lists six task types: data collection, model evaluation and feedback, annotation and tagging, linguistic editing, transcription and speech, and prompt engineering.
Whatever the task, the same habits apply: follow the project rules, expect your work to be checked, log your time honestly and keep project details private.
Sources: Mercor experts page, Lionbridge AI opportunities.
Follow the Project Rules
Each project comes with its own instructions. On micro1, once you are matched, a Human Data Manager gives you instructions, rubric guidance and support. Ethos requires a tutorial before any project, and you may need to complete it again each year.
micro1 says consistency, following the rubric precisely, is how quality is measured. Mercor asks experts to apply judgment, domain knowledge and professional standards.
Sources: micro1 guide, Ethos expert terms, Mercor experts page.
Expect Your Work to Be Checked
Every Alignerr project includes a calibration phase where sample work is reviewed, and quality and speed together form a score. Ethos sets quality standards for project work and may remove experts who fail to meet them.
Alignerr accounts may be suspended for policy violations, low-quality work or inactivity.
Sources: Alignerr FAQ, Ethos expert terms.
Log Your Time Honestly
Insightful is the time tracker Mercor requires on its projects, and it records hours worked, active and idle time, and app and website use.
Mercor reviews logs for each weekly pay period, and a human reviewer can remove idle time, time outside the project and time that breaks its AI usage policy. It bans inflating hours with low or simulated activity, and software that masks or manipulates the tracker. Mercor also caps weekly hours to keep quality up.
Sources: Mercor time tracking and pay policy, Mercor account security and conduct policy, Insightful.
Keep Project Work Private
Ethos's Expert Terms require strict confidentiality of non-public project information. That is why this guide covers habits, never the content of real tasks.
Mercor says projects likely to include sensitive material are flagged in advance, declining or stepping back from one never affects your standing, and live wellness support opens after 40 hours of project work.
Sources: Ethos expert terms, Mercor wellness resources.
Put these skills to work
Chat with DigiNo Toucan. A few quick questions, no CV needed.
Ready to apply? The matcher on our AI training jobs page points you to platforms that fit your background.
AI Training Tips FAQ
Is AI training work only rating AI answers?
No. Mercor describes it as reviewing AI answers in your area, creating examples that teach models how experts think, and checking quality against clear standards. Lionbridge lists six task types, from data collection to prompt engineering.
How is the quality of my work judged?
micro1 says consistency, following the rubric precisely, is how quality is measured. Every Alignerr project includes a calibration phase where sample work is reviewed.
Can logged hours be removed?
On Mercor, a human reviewer can remove idle time, time outside the project and time that breaks its AI usage policy.
Can I share the tasks I work on?
Ethos's Expert Terms require strict confidentiality of non-public project information, so keep task content private.
