Primary Care Meets Medical AI: A UK GP's Guide to Earning £50–100/hr

If you are a UK GP, you are sitting on the most commercially valuable clinical skill set in the AI training market: undifferentiated, multi-system, time-pressured decision-making across the entire breadth of medicine. AI companies cannot replicate that with hospital specialists alone — and industry reporting suggests they are paying physicians £50 to £160+ per hour to teach their models how primary care actually works.¹

This guide is for GP partners, salaried GPs, locum GPs, GP trainees (ST1–ST3), and recently retired GPs who want to understand what AI work looks like, what it pays, and how it fits around UK general practice.

Why GPs Are the Most Sought-After Clinicians in AI Training

The large language models powering medical AI products were initially trained on hospital-style content — textbooks, journal articles, specialist guidelines. The result is models that can recite the management of a STEMI but stumble on the realistic primary care presentation: the 53-year-old who mentions chest discomfort at the end of a 10-minute appointment about their daughter's eczema.

GPs bring four things AI companies cannot get anywhere else:

• Undifferentiated presentations — assessing symptoms before a diagnosis exists, not after
• Multi-morbidity reasoning — managing the patient with diabetes, COPD, depression, and chronic back pain in a single consultation
• NICE / NHS pathway fluency² — knowing what is actually commissioned, what triggers an urgent suspected cancer referral under the Faster Diagnosis Standard,³ and what gets bounced back from secondary care
• Safety-netting and risk stratification — the explicit "what to do if it gets worse" reasoning that hospital documentation rarely captures

This is precisely the content that AI products aimed at patients, triage tools, and clinician co-pilots are weakest at — and it is what AI companies are most willing to pay for.

What the Work Actually Looks Like for GPs

Most GP-suitable AI work falls into four buckets:

1. RLHF (Reinforcement Learning from Human Feedback) — you read AI-generated responses to clinical prompts and rate them, correct them, or rank two answers against each other. Typical prompt: "A 42-year-old woman presents with a 3-week history of fatigue and weight loss. What is your differential and initial workup?"
2. Clinical guideline annotation — labelling AI outputs against NICE, BNF, or NHS England pathways. Strong fit for GPs because primary care is where most NICE guidance actually lives.
3. Red-teaming and safety review — deliberately probing AI products to find unsafe advice (missed red flags, dangerous prescribing, inappropriate reassurance). High-paying, lower-volume work.
4. Clinical advisory and SME interviews — paid 30–60 minute calls with AI product teams explaining how primary care really works. Often £100–150/hour.

The work is asynchronous, remote, and you choose your own hours. There is no rota, no clinical responsibility, and no patient contact.

Earning Scenarios for UK GPs

The figures below are illustrative ranges based on publicly reported clinician rates from US-based AI training platforms (e.g. Mercor, Surge AI).¹ UK-specific rate cards are not published; actual offers vary by platform, specialty, project, and demand.

How AI Work Compares to the Alternatives GPs Already Know

Factor	AI Work	OOH Sessions	Private GP	Locum Sessions
Hourly rate (illustrative)	£50–110	£80–120	£100–200	£90–110
Clinical indemnity required	Check with your MDO4	Yes	Yes	Yes
Commute	None	Variable	Clinic-based	Variable
Clinical liability	None (no patient contact)	Full	Full	Full
Anti-social hours	None (you choose)	Yes	Sometimes	Sometimes
Cancellation risk	Low	Low	Medium	High
CPD value	Emerging	Established	Established	Established
Set-up time	15 mins + verification	Onboarding pack	Significant	Onboarding pack

The most common decision GPs make is not "AI work instead of clinical practice" — it is "AI work instead of OOH or locum sessions". The arithmetic is straightforward: comparable headline pay, no commute, no risk of being the responsible clinician for a patient you have never met.

GMC and Contractual Considerations for GPs

There are three areas to think about. None are blockers, but all are worth getting right.

1. GMC Good Medical Practice (2024) — does not prohibit secondary employment. Paragraph 95 requires that any conflicts of interest are declared and managed.⁵ AI work is remote, asynchronous, and involves no patient contact, so the typical GMC concerns (fitness to practise, fatigue, patient confidentiality) rarely apply. Including AI work in your appraisal portfolio for revalidation is generally a positive — it demonstrates engagement with emerging clinical technology.⁶

2. Your contract

• Salaried GPs: check the secondary employment clause in your BMA model salaried GP contract.⁷ Most practices require notification, not permission.
• GP partners: check your partnership deed. Partnership agreements typically include clauses on outside earnings — the BMA's partnership deed guidance is a good starting point.⁸
• Locum GPs: see the BMA's sessional and locum GP contract guidance.⁹ No contractual restrictions in either direction.
• GP trainees (ST1–ST3):¹⁰ check your deanery's secondary employment policy. AI work is generally permitted, and combined hours must remain compliant with the European Working Time Directive (48-hour weekly average).¹¹

3. Tax — AI income is self-employed income. You will need to register for Self Assessment if you have not already,¹² and most GPs will operate as sole traders initially. See our full GMC and tax guide for IR35,¹³ pension, and limited company considerations.

Why Verification Matters — and Why Generic Platforms Fail GPs

The dominant AI training platforms (Outlier,¹⁵ Mercor,¹⁶ Scale AI, Surge AI) treat all clinicians as broadly interchangeable. A platform asks "are you a doctor?" — you tick yes — and you are placed in the same pool as international clinicians, residents, and anyone else who claimed a medical degree.

This causes three problems for UK GPs:

• You compete on price with global clinicians rather than being matched to UK-specific work that needs UK-specific expertise.
• You are not surfaced for the highest-paying work (NICE-referenced annotation, NHS pathway review, MHRA safety work) because the platform cannot prove your registration.
• AI companies looking specifically for GMC-registered GPs cannot find you in a sea of unverified profiles.

EnterTheLoop is built around the opposite premise: every clinician is GMC-verified against the public register¹⁷ before being matched to roles. AI companies pay a premium for that verification because it removes their compliance risk — and that premium is reflected in your hourly rate.

The healthcare AI market is large and growing fast: Grand View Research projects it to reach approximately $187.7 billion by 2030,¹⁸ and demand for verified UK clinicians (required by NHS and MHRA-regulated AI products) is growing faster than supply.

Getting Started as a GP

The path from "interested" to "earning" is straightforward:

1. Register on EnterTheLoop — select "Doctor" and specify "General Practice" as your specialty
2. Add your training grade (GP partner, salaried GP, locum, ST1–ST3, retired) and special interests (e.g. women's health, dermatology, mental health)
3. Upload your credentials — GMC certificate, photo ID, evidence of CCT or training number
4. Get GMC-verified — we check your registration against the public register (2–3 business days)
5. Get matched — receive AI roles matched to your specialty, special interests, and availability

GPs with declared special interests typically receive 2–3× more role matches in the first month than those without, because most AI annotation work is sub-specialty scoped (e.g. "verified UK GPs with women's health experience").

FAQ

Can GP trainees (ST1–ST3) do AI work?

Yes. AI work is permitted by most deaneries provided it does not interfere with training or breach the European Working Time Directive when combined with clinical hours. It is not counted towards training hours. Including it in your portfolio demonstrates engagement with emerging clinical technology.

Will AI work affect my CCT or revalidation?

No. CCT requires completion of the GP training curriculum — AI work neither helps nor hinders this. For revalidation, AI work is a legitimate scope of practice to declare, and the reflective learning generated (clinical reasoning, exposure to edge cases) can support your appraisal.

Can retired GPs still do AI work?

Yes — and retired GPs are highly sought-after. Many AI companies specifically want clinicians with decades of UK general practice experience. You need an active GMC registration (with a licence to practise is preferable but not always required for non-clinical advisory work).

Does AI work count as private practice for indemnity purposes?

Probably not, because there is no patient contact and no clinical decision-making affecting a real patient — but neither the MDU nor MPS has published explicit guidance on RLHF-style work, so confirm scope with your medical defence organisation before starting. Most contracts also include their own indemnity clause from the AI company.⁴

How does AI work fit around a 6 or 8 session week?

The most common pattern is 1–2 evenings per week (5–8 hours) or one weekend morning (4 hours). Because the work is asynchronous, you can stop and start mid-task — useful for GPs whose evenings are unpredictable.

What if I have a specialty interest (e.g. dermatology, women's health, paediatrics)?

Specialty interests significantly increase your earning potential. AI companies actively pay premium rates for sub-specialty annotation — verified UK GPs with declared interests in women's health, mental health, dermatology, MSK, and paediatrics are in particularly high demand.

How quickly can I start earning?

Most GPs receive their first role match within 1–2 weeks of completing GMC verification. From sign-up to first payment is typically 3–4 weeks.

Is this a fad?

Independent forecasts put the healthcare AI market at $110bn–$188bn by 2030.¹⁸¹⁹ UK-specific AI products (NHS-deployed triage, patient-facing chatbots, clinician co-pilots) require UK-trained clinicians by regulatory necessity, not preference. Demand for verified UK GPs is growing faster than supply.

---

Sources & References

Footnotes

1. Mercor primary care physician rates and Surge AI medical fellow rates reported by CNBC (CNBC, Dec 2025) and SF Standard (April 2026). UK-specific rate cards are not published; figures here are illustrative. ↩ ↩²

2. National Institute for Health and Care Excellence (NICE) — About NICE guidelines. ↩

3. NHS England — National Cancer Waiting Times Monitoring Dataset Guidance: Faster Diagnosis Standard (28 days). The 2-week-wait target was replaced in October 2023. ↩

4. The Medical Defence Union (themdu.com) and Medical Protection Society (medicalprotection.org/uk) have not published explicit guidance on RLHF/AI training work — confirm scope of cover directly with your MDO before starting. ↩ ↩²

5. General Medical Council — Good Medical Practice (2024), paragraph 95 on declaring conflicts of interest. ↩

6. General Medical Council — Revalidation requirements for doctors. ↩

7. British Medical Association — Salaried GP model contract toolkit. ↩

8. British Medical Association — The importance of an up-to-date GP partnership agreement. ↩

9. British Medical Association — Sessional and locum GP contract guidance. ↩

10. Royal College of General Practitioners — CCT and GP training pathway. ↩

11. British Medical Association — Doctors and the European Working Time Directive. ↩

12. HMRC — Register for Self Assessment. ↩

13. HMRC — Understanding off-payroll working (IR35). ↩

14. HMRC — Tax on your private pension contributions: annual allowance. Allowance raised from £40,000 to £60,000 in April 2023. ↩

15. Outlier — Medical Expert page. ↩

16. Mercor — Marketplace. ↩

17. General Medical Council — The Medical Register. ↩

18. Grand View Research — AI in Healthcare Market Size & Trends (forecast ~$187.7bn by 2030). ↩ ↩²

19. MarketsandMarkets — Artificial Intelligence in Healthcare Market (forecast ~$110.6bn by 2030). ↩