Platform review
Turing
Real frontier-lab projects behind a long unpaid screening pipeline, and almost no published pay rates.
Turing recruits credentialed specialists for frontier-lab data work, and it is not a task queue. Its postings are specific about what you will produce, from authoring scientific problems to evaluating model output in your own field, and specific about the standards the output gets judged against. Getting there means a form, skill assessments, an interview and identity verification, all unpaid, over a window Turing puts at “48 hours to 6 weeks”. What Turing mostly will not tell you is what any of it pays: most postings on its own board carry no rate, and the dollar figures on the role cards are referral bonuses rather than wages. It suits a specialist who can afford to spend unpaid hours on a contract that might pay well, and it does not suit anyone who needs the number first.
Key facts about Turing
- Pay
- Most roles publish no rate
- Application
- Interview / vetted
- Availability
- Global
- Specialties
Read this first
The things that change how you should read everything else on this page.
- Keep the two Turing businesses apart. Most of Turing's Glassdoor and Indeed reviews are about developer placement and internal staff jobs, not the AI advancement work. Ratings drawn from the placement side are not ratings of AI training work, and anyone quoting the 4.7 or the 3.5 at you is almost certainly making that substitution.
- Do not read the numbers on the job cards as pay. Where a role card at work.turing.com or on Turing's talent page shows a dollar figure, and we saw $100, $150, $200, $250, $300 and $1,000, it is the referral bonus: open the card and it reads “Referral Bonus” beside “Share & earn”. Many cards show no figure at all.
- Turing's talent page advertises a 4.3 Trustpilot score. The turing.com Trustpilot profile read 2.5 out of 5 on September 12, 2026.
- The aggregator hourly ranges do not agree with each other or with Turing's own postings. Glassdoor showed $200 to $300 an hour on Turing legal and software listings, Indeed showed $100 to $150 an hour on Turing expert listings, and Turing's own Legal Expert posting states $400 per approved task with no handling time given. None of those is a figure a contributor has been reported as receiving.
- The allegation in recent Trustpilot reviews that Turing keeps unpaid test answers as training data is an unverified claim made by rejected candidates. It is reported here as a dated allegation, not a finding.
- Indeed's 4.7 sits on a stated 89 reviews while the page header says 67, and the visible reviews run out in January 2025. It is weak evidence about contractor experience in 2026.
- Turing restructured its help center and its older Zendesk-era articles now return 404s, including the one that previously described the project payment and payout process. Expect help center links to move.
What the work is
Turing is a San Francisco company running two businesses, and only one of them is what a reader of this site is after. On its jobs page it splits the work into application engineering, which deploys AI systems into enterprise applications like copilots and workflow automation, and AI advancement, which uses industry, technical or academic expertise to improve foundation models' accuracy and reasoning; the homepage gives the same split the customer-facing names Enterprise AI and Frontier AI, and the enterprise arm carries the Turing Intelligence brand. AI training work sits on the AI advancement side, and means writing and validating problems, evaluating model outputs, and reviewing generated content in your own field. Turing's homepage claims 300+ RL environments, 5M+ domain experts, 1M+ curated tasks and partnerships with “9/9 frontier labs”; those are company figures.
Who they want
Turing's talent page says it recruits professionals across tech, finance, healthcare, law and creative fields from 140+ countries, and its testimonial reel leans hard on academic credentials: of six experts profiled, three are introduced by graduate degree (MS NYU, PhD UIUC, PhD UCSF) and two more by institution (UC Berkeley, Boston University). The postings match that framing. Scientific coding roles require a master's or PhD in the subject plus strong Python and scientific computing experience; the Legal, Risk and Compliance roles want a JD with active US bar admission and three to eight years of practice; the registered nurse posting wants a BSN, four years of clinical experience and an active RN license. Individual postings are often geo-restricted, sometimes tightly, and in opposite directions: the biology scientific coding role is open only to applicants in Bangladesh, Brazil, Colombia, Egypt, Ghana, India, Pakistan, Indonesia, Kenya, Nigeria, Turkey and Vietnam, while the legal roles require US bar admission.
Step 1
Apply on the job board
A short form at work.turing.com/jobs that Turing says takes under two minutes.
Step 2
Proctored assessment
Some roles only: desktop, no private browsing, and your entire screen shared.
Step 3
Interview for senior roles
Senior postings add a live or AI video interview on top of the assessment.
Step 4
Review, 48 hours to 6 weeks
No feedback on a rejection, no withdrawal, and three months before you reapply.
Step 5
Identity check, then the wait
Onboarding verifies identity; work arrives by email or WhatsApp, on no stated timetable.
How you get in
You apply from work.turing.com/jobs, which carries the live board and a short form Turing says takes under two minutes. Some roles then drop you straight into a proctored assessment that requires a desktop computer, rules out incognito and private browsing, and makes you share your entire screen rather than a single window or tab; senior roles add a live or AI video interview, and at least one states the application alone takes 15 to 30 minutes. Turing puts application review at “48 hours to 6 weeks” and is blunt about the rest: not every candidate who is not selected will hear back, no individualized feedback is given on a rejection, an application cannot be withdrawn once submitted, and a formal rejection carries a three-month wait before you can reapply to the same role. If you get through, onboarding includes identity verification, and new opportunities arrive by email or WhatsApp. Two things Turing documents nowhere, and they are the two that decide whether the pipeline is worth entering: how many assessment hours a role costs you before anyone decides, and how long the wait runs between being accepted and receiving a first task.
How it pays
No open roles to plot. We are not mirroring any open role on Turing's board right now, so there is no distribution to draw.
Most Turing postings state no rate at all. The dollar figures on the role cards at work.turing.com and on the talent page are referral bonuses rather than pay: open the $100 scientific coding card and it reads “Referral Bonus · $100” beside “Share & earn $100”, and many cards carry no figure whatsoever. A few postings do publish a rate and repay close reading: the Legal Expert role in Legal, Risk and Compliance is “Pay Per Task” at “$400/approved task”, with the average-handling-time field left blank, so there is no way to convert it to an hourly figure, and the word “approved” carries real weight next to the Tier 1 and Tier 2 quality checks the scientific coding roles describe. Turing's listings syndicated to job aggregators show wider and mutually inconsistent numbers: on September 12, 2026 Glassdoor's Turing jobs panel displayed $200 to $300 an hour marked employer-provided on legal and software roles, while Indeed's panel showed $100 to $150 an hour on five remote expert roles. On cadence, the talent page promises “Transparent, bi-weekly payouts”, and Turing's one live payments help article covers referral payouts only, putting those bi-weekly by bank transfer with third parties possibly used for processing and contractors responsible for their own taxes; Self-Made Success reported in June 2026 that Turing “appears to pay monthly” via Payoneer, Wise or direct deposit, and worker accounts describe project payments running through Deel.
What workers say
The public record on Turing
The evidence on Turing is plentiful but poorly matched to the question an AI trainer is asking. Trustpilot's turing.com profile is the closest fit and it is harsh: 2.5 out of 5 across 201 reviews, and the recent ones are dominated by people reviewing the application pipeline they had just been through rather than the job, with unpaid assessments, failed identity verification and stuck payments the recurring themes. Glassdoor (3.5 from 783 reviews) and Indeed (4.7 from a stated 89) are far kinder, but they measure employment at Turing broadly; of Glassdoor's top reviewed job titles only LLM Trainer, at 41 reviews, matches AI training work, and Indeed's visible reviews stop in January 2025 and come almost entirely from internal staff and placed engineers. Complaints cluster hard around the pipeline, getting in, getting verified, getting assigned and getting paid, while the people who reach actual tasks tend to describe the work itself well.
Scores, and what each one measures
- Trustpilot (turing.com) ↗2.5 out of 5
What it measures. Mostly the recruiting and payment pipeline, not the job. The recent reviews driving the score come from candidates describing assessments that led nowhere, identity verification that failed, referral payouts that did not match the dashboard, and withdrawals that sat unresolved. Very few are from people who were actually doing AI training tasks. The 67% five-star share sits against a 2.5 score because Trustpilot's TrustScore weights recent reviews more heavily; an unweighted average of that distribution would land near 3.8.
- Glassdoor (Turing, E2462330) ↗3.5 out of 5; 61% would recommend to a friend; 69% approve of the CEO; 54% positive business outlook. Category ratings published in the page's FAQ block: compensation and benefits 3.6, work-life balance 3.7, culture and values 3.2, career opportunities 3.1.
What it measures. Employment at Turing across both business lines, weighted heavily toward placed engineers and internal staff. The top reviewed job titles are Software Engineer (101), Data Scientist (81), Business Analyst (74), Senior Software Engineer (52) and Python Developer (51); LLM Trainer accounts for 41 of the 783 and is the only title that clearly maps to AI training work. Glassdoor displays a six-month ratings trend chart but publishes no monthly figures behind it, and most review bodies require a sign-in.
- Indeed (turing) ↗4.7 out of 5 (work-life balance 4.7, pay and benefits 4.4, job security and advancement 4.5, management 4.6, culture 4.7)
What it measures. Internal staff and placed engineer roles from 2022 to January 2025, not AI training work. Visible reviewers include a delivery data analyst, senior product designer, senior director, account executive, front end engineer, data specialist, data scientist, pre-sales staff and a developer support specialist. A large block of the 2022 reviews follow a near-identical first-person format opening with the writer's home country and praising flexibility and culture, which reads as a solicited review drive rather than independent feedback.
Our read of it
Pros
- The postings describe the work in specific, checkable terms. The scientific coding biology role states a partnership with NVIDIA and asks for one main problem plus at least three linked sub-problems, reference solutions in Python with full unit test coverage, and test cases designed to separate correct model output from incorrect, scored against GPT, Gemini and Nemotron judges. You know what you are expected to produce before you apply.
- The board is genuinely deep. On September 12, 2026 work.turing.com listed 217 roles across ten categories including Science, Healthcare, Legal, Linguistics and Finance, so a specialist outside software has somewhere to apply.
- A handful of roles publish a real number. The Legal Expert posting's $400 per approved task is a figure a lawyer can actually evaluate, which is more than most platforms offer professionals at that level.
- Initial intake can be quick. A Trustpilot reviewer in August 2025 described the portal auto-populating a profile from an uploaded resume with about a minute of review needed, and Turing says its application form takes under two minutes.
- Contributors outside the US rate the pay well against local options. A Ghana-based reviewer on Indeed set Turing's flexibility, culture and timely pay against “limited options and low pay” for developers locally, and Glassdoor reviewers rate Turing 3.6 out of 5 on compensation and benefits. Both of those groups are mostly placed engineers and internal staff rather than AI training contributors.
- Remote and flexible is the most repeated positive across Glassdoor and Indeed, including in the newest Glassdoor reviews. A September 1, 2026 reviewer named flexible timing and fully remote work as the job's best features.
Cons
- Unpaid assessment time that ends in nothing is the single most repeated complaint. Trustpilot reviewers through August and September 2026 describe completing coding and language tests, sometimes over several rounds, and receiving neither an offer nor an explanation. Several go further and allege Turing is keeping their test answers as training data. Turing has not addressed that allegation either way.
- Most postings will not tell you what they pay, and the ones that do tie pay to approval. The scientific coding roles ask for 40 hours a week, four hours of PST overlap and an eight-week contractor assignment with no medical or paid leave, and state no rate; the registered nurse posting offers only “Strong compensation (exact amount varies by project)”. Where a rate does appear it is per approved task, and Turing publishes nothing about what happens to your pay when work fails quality control or goes back for rework.
- Identity verification blocks people at the last step. A reviewer describing an August 18, 2026 experience said the ID capture tool auto-cropped their documents, that both automated attempts and the manual review then failed, and that the final rejection stated no further information would be provided. That reviewer quotes a published Turing policy promising rejected candidates specific reasons for rejection, which is the opposite of what they describe happening.
- Money gets stuck, and the referral dashboard is not reliable. One Trustpilot reviewer reported a withdrawal request unresolved from June 8 to August 8, 2026 despite repeated escalation, and another described payment frozen in a blocked Deel account. In an August 2026 review a contributor said his dashboard showed $1,450 eligible, support paid a fraction, and Turing replied that the dashboard may differ from its internal records, then applied a newer referral policy to older referrals.
- Being accepted is not the same as getting work. A July 2026 reviewer described passing onboarding in May, having their profile locked to “assigned to a client”, receiving no tasks for two months, and being blocked from applying to other roles the whole time.
- Support is heavily automated and Turing barely replies. Recent Trustpilot reviews describe automated loops with no reachable human reviewer, while recruiting messages on WhatsApp and LinkedIn keep arriving after a rejection. Trustpilot's own profile data says Turing has replied to 12% of negative reviews and typically takes over a month to do so.
Who it fits, and who should skip it
Best for
- PhDs and master's holders in a STEM field with strong Python
- Licensed professionals: lawyers with US bar admission, clinicians, accountants
- Specialists outside software who want frontier-lab project work
- People who can absorb unpaid screening hours without needing the money soon
Who should skip it
Skip Turing if you need income soon or cannot absorb unpaid time. Review runs from 48 hours to six weeks, most applicants hear nothing, a rejection comes with no feedback and a three-month wait before you can reapply to the same role, and most postings state no rate to weigh any of that against. Skip it too if you want light, pick-up-and-drop task work, though check the individual posting first, because commitments vary: the scientific coding roles want 40 hours a week with four hours of PST overlap on an eight-week term, while the legal roles are paid per approved task, and what is consistent across both is a contractor assignment with a fixed end date rather than an always-on queue.
Provenance
Where all of this came from
Every source behind this review
What each one says, and how much weight it deserves.
Names the two business lines Frontier AI (“Train and improve models”) and Enterprise AI (“Deploy, manage, and scale agents”), and carries the headline figures 300+ RL environments, 5M+ domain experts, 1M+ curated tasks and 9/9 frontier labs.
How to read it. Marketing page. Every figure on it is a company claim with no published basis.
Source for “Transparent, bi-weekly payouts”, the 140+ countries claim, $300M+ raised, the six credentialed testimonials, and an advertised 4.3 Trustpilot score. Its role cards carry dollar badges ($150, $200, $250, $300, $1,000) that are referral bonuses.
How to read it. Recruiting page. The advertised 4.3 Trustpilot score conflicts with the 2.5 on the turing.com Trustpilot profile.
Publishes the split this review is built on: application engineering, to “Deploy the world's most advanced AI systems into real-world applications”, and AI advancement, to “help foundational AI companies improve their model's accuracy and reasoning”. Lists the four-step process of profile, matching, assessments and live interviews, then onboarding.
How to read it. No pay figure appears anywhere on the page; it refers to “well-paid jobs” and “competitive pay” without numbers.
217 open roles on September 12, 2026 across ten categories (Software, ML Data & AI, Science, Business, Finance, Healthcare, Linguistics, Legal, Cloud, Other). The Scientific Coding biology posting states the NVIDIA partnership, an MS or PhD requirement, 40 hrs/week, four-hour PST overlap, an eight-week “Contractor assignment/freelancer (no medical/paid leave)” term and a 12-country restriction, with “Referral Bonus · $100” and no pay rate. The Legal Expert posting states “Pay Per Task” at “$400/approved task” with the average handling time field left blank.
How to read it. JavaScript-rendered, so the listings do not appear to a plain fetch. Role counts and individual postings change frequently.
Shows the current collection structure: About us, How to, Working with Turing, EDU Arena, Referral Program, Talent Onboarding and Support, with “How to get paid” among the popular articles.
How to read it. No article on identity verification, failed verification or account suspension appears in the collections or popular articles, and searches surface only retired Zendesk URLs that 404.
The documented application flow: apply at work.turing.com/jobs, a form under two minutes, proctored assessments requiring a desktop computer, full-screen sharing and no incognito mode, “Application reviews typically take between 48 hours to 6 weeks”, no notification guaranteed if you are not selected, no individualized feedback, applications cannot be withdrawn once submitted, a three-month wait before reapplying after rejection, and notification by email or WhatsApp.
How to read it. Stamped “Last Updated: 18 days ago” as of September 12, 2026. Turing's description of its own process, not an independent account of how it runs in practice.
The only live payments article. It covers the referral program: eligibility once a referral is hired and completes 10 working hours, bi-weekly payment by Turing's finance team, bank transfer as a supported method, third parties possibly used for payment processing, and contractors responsible for their own taxes.
How to read it. Referral payouts only. It says nothing about cadence, processor or currency for project work, and names no currency at all.
A healthcare AI training role requiring a BSN, four years of clinical experience and an active RN license, offering only “Strong compensation (exact amount varies by project)” and a roughly one-month contractor assignment.
How to read it. An individual posting that may be filled or edited. Turing's syndicated version of similar roles on Indeed showed $100 to $150 an hour, a figure absent from Turing's own page.
2.5 out of 5 from 201 reviews, 28 in the last 12 months, distribution 67/4/under 1/3/25, Turing replying to 12% of negative reviews and typically taking over a month. Recent reviews carry the dated accounts of unpaid assessments, the August 18, 2026 identity verification failure, the June 8 to August 8, 2026 withdrawal, the $1,450 referral dashboard discrepancy and the July 2026 bench account.
How to read it. Self-selected reviewers, skewed toward people with a grievance, and Trustpilot flags “No recent history of asking for reviews”. The aggregate score and the star distribution point in different directions because of recency weighting.
3.5 out of 5 from 783 reviews, 61% would recommend, 69% approve of the CEO, 54% positive business outlook, and FAQ-published category ratings of 3.6 compensation and benefits, 3.7 work-life balance, 3.2 culture and values, 3.1 career opportunities. Top reviewed titles include Software Engineer (101) and LLM Trainer (41).
How to read it. Mostly placed engineers and internal staff rather than AI training contributors. Most review bodies and the sub-rating panel require a sign-in, and the six-month trend chart publishes no monthly values. Individual Glassdoor job listing pages sit behind a bot challenge.
4.7 out of 5 with sub-ratings of 4.7 work-life balance, 4.4 pay and benefits, 4.5 job security and advancement, 4.6 management and 4.7 culture. The Ghana-based reviewer setting Turing's timely pay against “limited options and low pay” locally comes from here.
How to read it. The stated 89 reviews conflicts with the 67 in the page header, visible reviews stop in January 2025, and a large 2022 block follows a near-identical promotional format. Weak evidence about contractor experience.
Dated June 19, 2026 and scores Turing around 49%. Reports payment methods of “Payoneer, Wise, Direct deposit in some regions” and that Turing “appears to pay monthly”, and independently finds that many Turing listings display no compensation at all.
How to read it. Its payment mechanics are unsourced and its monthly cadence conflicts with Turing's own “bi-weekly payouts” claim. Its finding that listings carry no compensation matches what we read on the board.
Evidence scorecard
What we can back up about Turing
12
0
Not one of them is work we've done ourselves. Read every line below as what somebody else states, discounted by its tier.
- Tier 1 · Verified first-hand0 accounts
We did it ourselves and can describe it, with a date. No dated first-hand account of Turing's own process is on record here.
- Tier 2 · Platform-published8 statements
Stated on the platform's own site, board, or in writing to us. Read off Turing's own pages, and cited to them.
- Tier 3 · Worker reports0 reports
Said by people who say they've done the work. We hold none for Turing.
- Tier 4 · Unverified claim4 claims
Printed by a marketing or aggregator page with no sourcing behind it. We repeat these labeled so you can recognize them elsewhere, never as fact.
Listed pay
Board not mirrored yetNo listings mirroredWe don't sync Turing's own board, so there's no listed rate here for us to count.
Effective pay
Not knownNobody has published itWhat a person clears per hour of their life after unpaid application time, unpaid training, idle time between tasks and self-employment tax. Nobody publishes it for Turing, and we can't tell you yet either.
Getting in
Interview / vettedTier 2 · Platform-publishedTuring's own material describes the step, and the review names the page it came from. What share of applicants clear it, Turing does not publish.
Specialty labels
Board not mirrored yetNo listings mirroredSpecialties are read off listings, and we don't sync Turing's.