The best AI jobs right now fall into two groups. Jobs that train and check AI, like AI trainer and data annotator, are remote, paid by the hour, and often open to people without a degree. Jobs that build AI, like AI engineer and machine learning engineer, usually want a degree and pay six figures.
AI skills pay. PwC's 2026 Global AI Jobs Barometer, built from more than a billion job ads in 27 countries and territories, found workers with AI skills earn a 62% wage premium on average.
This guide covers 10 AI jobs, five non-technical and five technical: what each one is, what it pays, what you need to get hired, and where to find openings. I build AI tools for a living, and I went to each platform's own pay page, the US Bureau of Labor Statistics (BLS), and Levels.fyi so every number here has a source. Pay is in US dollars, because most of these jobs and platforms are American.
What are the best AI jobs in 2026?
The best AI jobs in 2026 are AI trainer, prompt engineer, AI content editor, AI product manager, and AI governance roles if you don't code, and AI engineer, machine learning engineer, data scientist, MLOps engineer, and research scientist if you do. Here's how the 10 AI jobs compare on pay, degree, and remote work.
| AI job | Typical US pay | Degree needed? | Remote? |
|---|---|---|---|
| AI trainer or data annotator | US$17 to US$50 an hour for general work, up to US$150+ for experts (platform rates) | No | Yes |
| Prompt engineer | US$133,000 a year (Glassdoor median) | Not always | Some |
| AI content editor | US$77,920 a year (BLS median for editors) | Usually | Some |
| AI product manager | US$230,000 a year (Levels.fyi, all product managers) | Usually | Some |
| AI ethics and governance | US$151,800 a year (IAPP median) | Usually | Some |
| AI engineer | US$151,000 a year (Levels.fyi) | Usually | Some |
| Machine learning engineer | US$280,000 a year (Levels.fyi) | Usually | Some |
| Data scientist | US$120,230 a year (BLS median) | Yes, a bachelor's | Some |
| MLOps engineer | US$161,000 a year (Glassdoor median) | Usually | Some |
| Research scientist | US$306,000 a year (Levels.fyi) | Yes, a master's or PhD | Some |
Levels.fyi figures are median total pay (salary, bonus, and stock) reported by people who work in tech, so they run higher than BLS medians, which cover everyone in that job across the US. AI training rates are each platform's own posted rates.
What are the best non-technical AI jobs?
The best non-technical AI jobs are AI trainer or data annotator, prompt engineer, AI content editor, AI product or operations roles, and AI ethics and policy roles. None of these AI jobs need you to write code, but all of them need you to know your way around AI tools.
AI trainer or data annotator
An AI trainer, also called a data annotator, rates, corrects, and writes examples for AI models so they give better answers. The work ranges from labelling images and ranking chatbot replies to writing hard maths, coding, or legal questions that trip the model up.
- Pay: About US$17 to US$50 an hour for general work and up to US$150 or more for coders and specialists, going by the platforms' own posted rates. DataAnnotation's FAQ lists US$25 to US$50+ an hour for general work and US$40 to US$150+ for coding.
- What you need: Strong written English, attention to detail, and a pass on the platform's entry test. Expert projects want a degree or professional experience in the field, such as law, medicine, finance, or software.
- Where to find openings: Apply directly on the AI training platforms: DataAnnotation, Outlier, Mercor, Alignerr, Handshake AI, and Prolific.
Prompt engineer
A prompt engineer writes and tests the instructions that steer an AI model inside a product, then builds test sets to check the answers hold up. It was one of the most wanted AI jobs a few years ago, and the standalone title has cooled since.
In Microsoft's survey of 31,000 workers in 31 countries, prompt engineering ranked second from the bottom of new roles companies were considering, and many firms are moving prompt engineers into other AI roles, Korn Ferry reports.
- Pay: Glassdoor's median total pay for prompt engineers is US$133,000 a year, according to Coursera's salary guide.
- What you need: Clear writing, a habit of testing prompts systematically, and examples of prompts and test sets you've built. Python helps, because the work increasingly sits inside AI engineering teams.
- Where to find openings: Search "prompt engineer", "LLM", and "AI engineer" on LinkedIn and Indeed. AI training platforms hire prompt writers too: Alignerr lists prompt writer work at US$40 to US$120 an hour.
AI content editor
An AI content editor fact-checks, edits, and rewrites AI-written drafts so they're accurate and sound human. The job shows up at publishers and marketing teams, and as writing evaluation work on AI training platforms.
- Pay: The BLS median for editors is US$77,920 a year, though the BLS expects editor jobs overall to shrink 1% over the decade to 2035. On AI training platforms, Alignerr lists English writing quality review at US$30 to US$40 an hour.
- What you need: A sharp editing eye and samples of your work. The BLS says editors typically have a bachelor's degree in communications, journalism, or English.
- Where to find openings: Content and marketing job ads that mention AI, and the writing projects on DataAnnotation and Alignerr.
AI product manager or AI operations
An AI product manager decides what an AI feature should do and who it's for, while AI operations roles roll AI tools out across a business and check they're working. AI sales and solutions jobs, which explain AI products to customers, sit in this group too.
- Pay: Levels.fyi's US median total pay for product managers is US$230,000 a year. Levels.fyi doesn't track AI product managers separately.
- What you need: Experience running products, projects, or operations, plus enough hands-on time with AI tools to know what they can and can't do. The easiest route is often an internal move: become the person on your team who rolls out the AI tools.
- Where to find openings: LinkedIn and Indeed searches for "AI product manager", "AI operations", or "AI solutions".
AI ethics and policy
AI ethics and policy jobs, now usually called AI governance, check AI systems for bias, privacy, and legal risk, and set the rules for how a company uses AI.
- Pay: In the IAPP's Salary and Jobs Report, half of people working only in AI governance earned more than US$151,800 a year. The survey covered more than 1,600 people in over 60 countries.
- What you need: A background in law, privacy, compliance, or policy, plus a working understanding of how AI models are built and tested. The IAPP's AIGP certification (Artificial Intelligence Governance Professional) is built for this job.
- Where to find openings: Legal, privacy, risk, and trust and safety teams at larger companies, plus government and policy organisations.
Which AI training jobs pay the most?
Mercor and Alignerr post the highest average pay for AI training jobs, at US$110 and US$80 an hour, because they mostly hire professionals and PhDs. For general AI training work without a specialist background, DataAnnotation and Handshake AI are the easier ways in.
| AI training platform | Posted pay | Who it hires | Who can apply |
|---|---|---|---|
| DataAnnotation | US$25 to US$50+ an hour general, US$40 to US$150+ coding | Generalists (no degree) and experts | Fluent English speakers; no country list published |
| Outlier | Not published; shown before each project | Associate degree or higher; experts in coding, STEM, and languages | ID and a mobile phone from your country of residence |
| Mercor | US$110 an hour average contracted rate | Lawyers, doctors, engineers, and consultants | Remote from anywhere; no country list published |
| Alignerr | US$80 an hour average; listings from US$20 to US$120 | PhDs, researchers, and experienced professionals | No country list published |
| Handshake AI | Top rates from US$17 to US$125 an hour, by role | Students, graduates, and professionals; no AI experience needed | US only, with work rights |
| Prolific | At least US$8 an hour (it recommends US$12) | Anyone accepted off the waitlist | Waitlist, then an ID check |
What each AI training platform is like:
- DataAnnotation is the closest thing to an open door. Its generalist role needs no degree, just strong reading, attention to detail, and clear written English. You only get one go at its Starter Assessment, and it pays by PayPal.
- Outlier is run by Scale AI. Its FAQ doesn't publish rates; you see the rate before each project starts, and pay goes out weekly by PayPal, Airtm, or bank transfer. You need at least an associate degree.
- Mercor hires professionals for expert projects after a short AI evaluation and an AI interview, according to its experts page. Recent listings include legal experts at US$90 an hour and healthcare experts at US$150.
- Alignerr is the expert platform from data labelling company Labelbox. Its listings run from US$20 to US$50 an hour for sales operations task writing up to US$40 to US$120 for prompt writing and data labelling specialists.
- Handshake AI is the AI training arm of the Handshake careers network. Its roles run from up to US$17 an hour for AI evaluation specialists to up to US$125 for game developers, and you must be based in the US.
- Prolific pays people to take part in research studies, including evaluating AI models. Its researcher rules set a minimum of £6 or US$8 an hour and recommend at least £9 or US$12.
Every rate here is the platform's own figure, and most are "up to" rates. AI training work comes in projects, so your hours depend on what's running. Sign up to two or three platforms so one empty queue doesn't leave you with nothing.
How do you avoid AI training job scams?
A real AI training job never asks you to pay. Scammers borrow the names and language of real AI training platforms, so if an "AI job" wants a fee, a deposit, or crypto to unlock tasks or withdraw earnings, it's a scam.
Job scams are a big business. Reported losses to job scams jumped from US$90 million in 2020 to US$501 million in 2024, according to the FTC. Reports of "task scams", which show you earnings for liking videos or rating product images and then demand a deposit before you can withdraw, went from about 5,000 in all of 2023 to about 20,000 in the first half of 2024 (FTC).
The red flags for AI training job scams:
- The offer arrives out of the blue by text, WhatsApp, or Telegram. The FTC says to ignore generic, unexpected job messages.
- You're asked to pay for training, equipment, or a deposit to "unlock" tasks or withdraw earnings. The FTC's rule is simple: never pay anyone to get paid.
- You're paid to "like" or rate things online and earnings grow by levels. That's the task scam pattern.
- The link doesn't go to the company's own site. DataAnnotation's trust and safety page says it only takes applications through its own sign-up flow and will "never ask you for money".
Are unpaid assessments for AI training jobs normal?
Yes, an unpaid entry test is normal for AI training jobs, but hours of unpaid work after you're in are not. DataAnnotation's Starter Assessment can only be taken once, Mercor screens with an AI interview, and Outlier says you're paid once you finish onboarding and start on a project.
The unpaid time doesn't always stop at the test. AlgorithmWatch interviewed more than 50 contractors for Outlier and its competitors and found training time and meetings at Outlier were mostly unpaid. One UK worker logged about 36 paid hours and 12 unpaid training hours in his first three months. A US worker put their real rate under US$5 an hour once unpaid time was counted.
Scale AI, which runs Outlier, has agreed to a US$12.5 million settlement of California class actions claiming its contributors were misclassified as contractors and underpaid. It denies wrongdoing.
To protect your time:
- Do the entry test, but don't do hours of unpaid "trial tasks" on top of it.
- Track your real hourly rate, counting unpaid training and time spent waiting for work.
- Never hand over bank details anywhere except inside the platform's own logged-in dashboard.
What are the best technical AI jobs?
The best technical AI jobs are AI engineer, machine learning engineer, data scientist, MLOps engineer, and research scientist. They pay the most of any AI jobs, and most want a degree in computer science, maths, statistics, or a related field, plus projects you can show. If you're starting from scratch, the best books on AI and machine learning are a good first step.
AI engineer
An AI engineer builds products on top of AI models: chatbots, search over company documents, agents that complete tasks, and the tests that check their answers.
- Pay: Levels.fyi's US median total pay for AI engineers is US$151,000 a year. Its separate applied AI engineer title has a median of US$250,000.
- What you need: Software engineering skills first, such as Python or TypeScript, APIs, and databases. Then hands-on work with large language models, including retrieval and evaluation. A computer science degree is common, but shipped AI projects carry a lot of weight.
- Where to find openings: LinkedIn, Indeed, and the startup job board Wellfound, searching "AI engineer", "LLM engineer", or "applied AI".
Machine learning engineer
A machine learning engineer trains, tunes, and deploys the models behind products like recommendations, fraud detection, and speech recognition.
- Pay: Levels.fyi's US median total pay for machine learning engineers is US$280,000 a year, on a median base salary of US$200,000.
- What you need: Python, a framework like PyTorch, solid maths and statistics, and the software engineering to run models in production. Most postings ask for a degree in computer science or a related field.
- Where to find openings: LinkedIn, Indeed, Wellfound, and the careers pages of the companies you want to work for.
Data scientist
A data scientist uses statistics and machine learning to find patterns in data and answer business questions, such as which customers are about to leave.
- Pay: The BLS median for data scientists is US$120,230 a year, and the BLS projects data scientist jobs to grow 35% over the decade to 2035. Levels.fyi's US median total pay for data scientists is US$180,000.
- What you need: Statistics, Python or R, and SQL. The BLS says data scientists typically need a bachelor's degree.
- Where to find openings: LinkedIn and Indeed, plus tech, finance, and health companies' careers pages.
MLOps engineer
An MLOps engineer builds the systems that train, deploy, and monitor machine learning models, so the models keep working after launch. Think of it as DevOps for AI.
- Pay: Glassdoor's median total pay for MLOps engineers is US$161,000 a year, according to Coursera's MLOps guide.
- What you need: A cloud platform (AWS, Azure, or Google Cloud), Docker and Kubernetes, Python, and enough machine learning to spot when a model's results start to drift. A DevOps or data engineering background is a common way in.
- Where to find openings: Search "MLOps", "ML platform engineer", and "ML infrastructure" on LinkedIn, Indeed, and Wellfound.
Research scientist
An AI research scientist invents new methods and models and publishes the results, at AI labs, big tech companies, and universities.
- Pay: Levels.fyi's US median total pay for research scientists is US$306,000 a year. The BLS median for computer and information research scientists, a broader group, is US$140,300, and the BLS projects those jobs to grow 22% over the decade to 2035.
- What you need: The BLS says research scientists typically need at least a master's degree. AI labs often want a PhD and published papers.
- Where to find openings: AI lab and big tech careers pages, university job boards, and the recruiting events at research conferences.
Where can you find remote AI jobs?
AI training platforms are the easiest place to find remote AI jobs, because all of the work is done online. DataAnnotation, Outlier, Mercor, Alignerr, and Handshake AI all hire remote contractors, and Mercor says you can work from anywhere you choose.
For technical remote AI jobs:
- Filter LinkedIn and Indeed by "Remote" and search the exact job title, such as "machine learning engineer", rather than just "AI".
- Use Wellfound for startup jobs, which you can filter to remote.
- Check Hacker News' monthly "Who is hiring?" thread, where many posts are marked REMOTE.
Outside the US? Handshake AI only takes people based in the US. DataAnnotation, Outlier, Mercor, and Alignerr don't publish a country list, so check when you sign up.
How do you get an entry-level AI job?
The quickest way into an entry-level AI job is AI training work, because it pays from day one, often needs no degree, and teaches you how AI models fail. From there, build towards the job you want.
- Start with AI training. Apply to DataAnnotation and one or two other platforms, and expect an unpaid entry test.
- Build something you can show. A small project, like a chatbot that answers questions from your own notes, says more than a certificate.
- Learn the basics. For the technical path, learn Python and work through one of the best books on AI and machine learning.
- Fix your resume. Lead with the AI tools you use and the projects you've built. The best AI resume builders can tailor it to each listing.
- Keep up with the field. One or two AI newsletters will show you new tools and new job types early.
If you'd rather work for clients than an employer, my guide to AI freelance jobs covers gigs you can start with AI tools, from AI video editing to chatbot builds.
