AI jobs

AI Jobs

The complete guide to AI jobs in 2026 — the two very different markets hiding behind the term, what AI engineers and AI training contributors actually earn, which platforms pay, what to check before signing up, and how to get in from wherever you are starting.

Updated August 2026

Quick answer: “AI jobs” describes two markets that have almost nothing in common, and knowing which one you are looking at is the single most useful thing on this page. The first is salaried employment building AI systems — AI engineer, machine learning engineer, MLOps, research — where US total compensation clusters around $170,000 to $245,000 at enterprises and runs to $600,000 or more at frontier labs, and where AI Engineer has been one of the fastest-growing roles on LinkedIn for three years running. The second is contract work training AI models — labelling data, writing prompts, grading model outputs — which requires no degree at some platforms, pays by the hour or by the task, and is accessible within days rather than years. The honest caveat that shapes everything below: pay in the second market is falling at the generalist end and rising sharply at the expert end, and the platforms that broker it offer no guaranteed volume — treat it as supplementary income rather than a salary.

This is the hub for our AI jobs coverage. Below, we separate the two markets, give the verifiable numbers for each, and link to the detailed page for every platform and route. Where a figure cannot be verified against a credible source, we say so rather than repeating the number everyone else is repeating — which, in this topic, is most of them.

The two kinds of AI job

Work out which row you are in before reading anything else. People routinely spend months preparing for the wrong one.

The two markets do connect, but weakly. Training work builds familiarity with how models behave and gives you something to talk about, and a minority of contributors move into prompt engineering, evaluation or data-operations roles. It is not a reliable pipeline into AI engineering, and any page telling you otherwise is selling something.

The state of the AI job market: August 2026

Four findings from primary sources, all of which cut against the prevailing narrative in one direction or another.

AI is adding jobs, and it is not what is slowing hiring. LinkedIn’s global labour market analysis, published by its Chief Operating Officer Dan Shapero on 15 January 2026, reports that the global economy added 1.3 million new AI-related roles in two years — AI engineers, forward-deployed engineers and data annotators — plus more than 600,000 AI-enabled data centre jobs (World Economic Forum). Global hiring is running about 20 per cent below pre-pandemic levels, but LinkedIn attributes that to economic uncertainty and monetary policy, not automation, and states the point directly: outside clinical healthcare roles, “hiring patterns look the same for jobs with high AI exposure and those with low exposure.”

AI job titles have escaped the tech sector. US job titles referencing AI more than tripled from 264 in 2022 to 822 by the first quarter of 2026, and 63 per cent of those titles now sit outside traditional technology occupations — in healthcare, education, marketing, logistics and management (Indeed Hiring Lab, 8 July 2026, reported by Network World). If you are not an engineer, this is the most important number on the page: the growth is in AI-adjacent roles inside ordinary industries.

Technology hiring is genuinely strong, against the headlines. Employers posted more than 280,000 new technology job postings in June 2026, a sixth consecutive month of growth, with active postings approaching 600,000 and tech-occupation unemployment at 2.9 per cent against a national rate of 4.2 per cent (CompTIA Tech Jobs Report, June 2026). Software developers and engineers led all categories with nearly 50,000 openings.

Entry-level is difficult but not collapsing. LinkedIn found the share of entry-level roles declined over three years but has “largely returned to historical norms,” while 52 per cent of people say they are job hunting in 2026 and nearly 80 per cent feel unprepared to do so. The competition is real; the door is not closed.

Path one: jobs building AI

The roles that are actually hiring

The last row is where most readers of this page will actually land, and it is the fastest-growing category. Nearly two-thirds of new AI job titles are outside technology occupations, and LinkedIn recorded a 70 per cent year-on-year increase in US roles requiring AI literacy. You do not need to become an engineer to hold an AI-titled job in 2026; you need domain expertise plus credible fluency.

What these roles pay

Compensation in AI engineering has split into two markets that share job titles but not pay scales.

At enterprises, AI and machine learning engineers cluster in the $170,000 to $245,000 total compensation range in the US, with machine learning engineer base salaries commonly reported between $128,000 and $186,000. At frontier labs, the same titles command $600,000 or more in median total compensation, with equity making up the majority of the package above mid-level seniority. Levels.fyi’s self-reported data puts median total compensation for machine learning engineers around $290,000 at Google and $457,000 at Meta; see AI salaries for the full company breakdown.

Treat all of these as self-reported aggregates, not audited figures — Levels.fyi data is submitted by employees, skews towards large well-paying employers, and is not a random sample. They are the best available reference points, not a guarantee.

How to get in

The realistic routes, in descending order of how well they work:

1. Move sideways from software engineering. The shortest path by a wide margin. Ship something that uses models in production, then apply into AI engineering.
2. Add AI to your existing domain. A nurse, teacher, logistics planner or marketer with genuine AI fluency is competing in a far thinner field than a junior engineer. This is where 63 per cent of new AI titles are.
3. Build and show work. Portfolios beat credentials in a market where, as LinkedIn puts it, “resumes become easier to embellish.”
4. Enter through the data layer. Data engineering and analytics roles are adjacent, established and hire more predictably than AI-titled roles.

If you do not have a degree, routes 2 to 4 are the realistic ones — none of them require going back to university.

Path two: jobs training AI

This is the market people mean when they search for AI work they can start this month. It is real, it pays, and it is more volatile than almost any coverage of it admits.

The platforms

The corporate relationships matter and are widely misreported. Scale AI is the parent contracting company; Outlier and Remotasks are both Scale-operated worker-facing platforms, with Outlier the higher-skill Western-facing front end and Remotasks the older, globally distributed microtask one. Outlier’s own FAQ states it is “a platform operated by Scale AI,” and its operating entity is Smart Ecosystem, Inc. Comparison pages that treat these as three competing companies are wrong.

What this work actually pays

There is no independent, methodologically sound survey of AI annotation pay for 2026, and most of the major platforms — Outlier, Surge, Appen and Remotasks among them — publish no rate card at all. What does exist falls into three tiers, and the tier matters more than the number: platform-advertised starting rates, named salary aggregators such as ZipRecruiter and Glassdoor, and vendor claims about their own networks. Our tier-by-tier ranges are in the AI training jobs and data annotation jobs guides, each labelled with its basis. Treat the ranges circulating on affiliate pages with no stated methodology, sample size or date as worthless — they contradict each other freely.

What can be verified:

The gap between the advertised averages and the individually reported rates is the story: a network average of $85 per hour sits alongside a documented content-review project cut from $21 to $16. Outlier states only that “rates vary by expertise, project complexity, and location” and that you will see the rate before accepting a project, which is a reasonable practice and also an admission that no standard rate exists.

Practical guidance: never judge a platform by an advertised average. Judge it by the rate shown on the specific project you are offered, and assume that rate can change or the project can end without notice.

The specialist shift is the defining trend

Generalist annotation work is being squeezed out while credentialed expert work commands multiples of it. This is one phenomenon, not two.

On the contracting side: xAI cut 500 roles from its data-annotation team in September 2025 — roughly a third of a 1,500-person team — and said it would prioritise specialist AI tutors over generalists, planning to expand the specialist team tenfold across STEM, finance, medicine and safety (TechCrunch, 13 September 2025). Scale AI cut 200 employees and 500 global contractors in July 2025, a month after Meta’s $14.3 billion investment for a 49 per cent stake, with interim CEO Jason Droege saying the company “ramped up our GenAI capacity too quickly” (Tom’s Hardware). In July 2026, Toloka folded its microtask contributor platform into Mindrift, its expert-oriented platform — a major microtask marketplace abolishing its own microtask front end.

On the demand side: OpenAI recruited more than 100 former Goldman Sachs, JPMorgan and Morgan Stanley bankers at $150 per hour to build financial models; Handshake lists $300 or more per hour for medical degrees and PhDs; Mercor positions explicitly on medicine, law and PhD-level reasoning and crossed $2 billion in annualised revenue in June 2026, reportedly in talks at a $20 billion valuation (TechCrunch, 9 July 2026).

Even entry requirements at the generalist platforms have hardened. Outlier now requires a minimum of an associate degree, with some projects requiring a bachelor’s, master’s or PhD.

What this means for you: if you hold a scarce credential — a medical or legal qualification, a PhD, professional finance or specialist engineering experience — this is the best-paid flexible work available to you right now, and you are being actively recruited. If you do not, expect competitive entry, falling rates and unstable project volume, and treat it as supplementary income rather than a salary.

Before you sign up: what to check

This market is lightly regulated and the terms are set by the platform, so a few checks are worth making before you commit time.

Confirm the rate on the specific project, not the platform average. Most of the major platforms publish no rate card, advertised averages are marketing, and rates vary by project, expertise and country.

Establish what is paid and what is not. Onboarding, qualification assessments, instruction-reading and time spent waiting for work are commonly unpaid. Divide the project rate by your realistic total time to get your true hourly rate.

Check the payment terms. Payment schedule, method, minimum thresholds, and what happens to a pending balance if your account is closed.

Check whether your country is eligible, and whether it is likely to stay that way — platforms have withdrawn from individual countries at short notice.

Never rely on one platform for essential income. Project volume is not guaranteed, projects end without notice, and accounts can be closed at the platform’s discretion. Contributors work as independent contractors, so employment protections do not apply.

Where to start, by situation

You want income within two weeks and have no specialist credential. Start with DataAnnotation.tech, which screens by assessment rather than résumé, and Outlier if you hold at least an associate degree. Read get paid to train AI first. Expect supplementary income, not a salary.

You hold a medical, legal, finance or PhD-level credential. You are the most sought-after category in this market. Mercor, Handshake AI and Surge AI all run credentialed tracks paying multiples of generalist rates. Verify the rate on the specific project before signing.

You want a career in AI and have engineering experience. Skip the training platforms entirely. Build something that runs in production, then apply into AI engineering — see “Path one” above.

You want a career in AI and do not have a degree. The honest routes are portfolio-led, data-layer-first, or domain-plus-AI rather than engineering-first — all three are in “Path one” above.

You are early-career and want a way in. The fastest-growing category is AI-titled roles inside non-tech industries, not junior engineering — add credible AI fluency to a domain you already know.

You want flexible or remote work specifically. See remote AI jobs and AI side hustles.

Frequently asked questions

What are AI jobs?

AI jobs fall into two distinct markets. The first is salaried employment building AI systems — AI engineer, machine learning engineer, MLOps engineer, AI researcher — which typically pays $170,000 to $245,000 in US total compensation at enterprises and considerably more at frontier labs. The second is contract work training AI models by labelling data, writing prompts and grading model outputs, which is paid hourly or per task and can be started within days. The two require completely different preparation and should not be researched together.

Do AI jobs pay well?

AI engineering roles pay very well: US total compensation commonly runs $170,000 to $245,000 at enterprises, and median total compensation at frontier labs reaches $600,000 or more, though those figures are self-reported aggregates rather than audited data. AI training contract work pays far less and is polarising: verified reporting shows credentialed specialists earning $150 to $300 or more per hour while generalist rates fall, including a documented cut from $21 to $16 per hour across more than 5,000 Mercor contractors in November 2025.

Can you get an AI job without a degree?

Yes for AI training contract work at some platforms, and yes but harder for AI engineering roles. DataAnnotation.tech screens by skills assessment with no résumé review or interview. Outlier, however, now requires a minimum of an associate degree, and some of its projects require a bachelor’s, master’s or PhD, so the no-degree route is narrowing even in gig work. For salaried AI roles, the realistic routes without a degree are portfolio-led applications, entering through data engineering, or adding AI fluency to existing domain expertise.

Is getting paid to train AI legitimate?

Yes. The work is real, the major platforms pay, and the clients are the AI labs whose models you are helping to train. What it is not is stable: contributors work as independent contractors, project volume is not guaranteed, projects end without notice, and accounts can be closed at the platform’s discretion. Verify the rate on the specific project before starting, work out your true hourly rate including unpaid onboarding, and never depend on a single platform as your only income.

Which AI training platform pays the most?

On verified evidence, the highest published rates go to credentialed specialists: Handshake AI has listed $300 or more per hour for contributors with medical degrees or PhDs and $175 per hour for investment bankers, and OpenAI paid $150 per hour to former bankers on its Project Mercury financial-modelling work. For generalist work, no reliable comparison exists because Outlier, Surge, Appen and Remotasks publish no rate cards, and the hourly figures circulating online come from affiliate sites with no stated methodology.

Is AI taking jobs or creating them?

Both, but the current evidence points more strongly to creation than destruction. LinkedIn’s January 2026 analysis found AI added 1.3 million new roles globally in two years plus more than 600,000 data centre jobs, and concluded that AI is not the cause of the wider hiring slowdown — outside clinical healthcare, hiring patterns look similar for jobs with high and low AI exposure. The clearest displacement inside the AI economy itself is generalist annotation work, where xAI cut 500 roles and Scale AI cut 500 contractors in 2025.

What is the fastest-growing AI job?

AI engineer has been among the fastest-growing roles on LinkedIn for three consecutive years. Structurally, though, the fastest-growing category is AI-titled roles outside technology: US job titles referencing AI more than tripled from 264 to 822 between 2022 and the first quarter of 2026, and 63 per cent of those titles now sit in healthcare, education, marketing, logistics and management rather than in software or data roles.

Are AI jobs remote?

AI training contract work is almost entirely remote, which is its main appeal, though location restrictions are real and enforced, and platforms have withdrawn from individual countries at short notice. Salaried AI engineering roles follow normal tech-industry hybrid patterns, with frontier labs generally expecting significant in-office presence. See remote AI jobs for the detail.

Is Outlier AI legit?

Outlier is a genuine platform operated by Scale AI, it pays weekly by PayPal, Airtm or ACH, and it shows the rate before you accept a project. Onboarding takes 30 to 90 minutes and now requires a minimum of an associate degree, with some projects requiring a bachelor’s, master’s or PhD. As with every platform here, onboarding and assessments are unpaid, so calculate your effective rate including that time.

How much do AI trainers make per hour?

There is no trustworthy answer to this question, and anyone giving you a confident range is guessing. No independent, methodologically sound survey of AI annotation pay exists for 2026, and the major platforms publish no rate cards. The verified individual datapoints span an enormous range: $16 per hour on a Mercor content-review project after a 24 per cent cut, around $31 per hour as a US self-reported average, $75 per hour for software engineers at Handshake, and $150 to $300 or more per hour for bankers, doctors and PhDs.

Explore AI jobs

Earning guides: Get paid to train AI · AI training jobs · Data annotation jobs · Remote AI jobs · AI side hustles · Prompt engineer salary

Platform reviews: DataAnnotation review · Outlier review

Build AI fluency: Best AI models · Best AI apps · Best AI for coding

Figures verified as of 15 August 2026. Compensation aggregates from Levels.fyi and similar sources are self-reported and unaudited; AI training platform pay figures are individually sourced and dated because no independent survey of this market exists. Platform terms, rates and country eligibility change frequently — confirm current details with the platform before signing up.