AI skills for doctors are no longer a side interest. In Doximity’s 2026 Physician Compensation Report, two thirds of physicians, 67 percent, said that staying current with AI tools will bring a meaningful earnings advantage over colleagues who do not adopt them during the next 12 months. The report, released on 25 August 2026, also found that 39 percent of physicians said AI proficiency already factors into hiring and promotion decisions in their specialty.
That is the short answer. The longer answer matters more for doctors who are building a career across systems, especially international medical graduates. AI fluency will not replace clinical judgement. It is becoming a form of professional currency because it changes how fast you can document, search, analyse and improve the service around you, and hiring teams are starting to notice.
What did Doximity actually find in 2026?
Doximity is a United States based digital platform for medical professionals. Its 2026 Physician Compensation Report drew on salary data from more than 23,000 physician responses collected in 2025, alongside a June 2026 survey focused on how AI is changing physician careers and pay.
Three findings stand out for any doctor weighing whether to invest time in AI.
First, use is already mainstream. Around two thirds of physicians, 66 percent, reported using AI on a daily or weekly basis for clinical or administrative work. Adoption was highest among physicians in their thirties, at 73 percent using AI at least weekly, compared with 52 percent among physicians in their sixties.
Second, doctors do not mostly see replacement. Nearly three quarters, 72 percent, said AI is not changing the physician role in their specialty, while half, 50 percent, said it is changing how the work gets done. In other words, the job title remains, the workflow does not.
Third, the career effect is arriving early. Alongside the 67 percent earnings advantage figure, 23 percent of physicians expected AI to increase their total compensation within the next 12 months. At least 39 percent said AI proficiency is a factor in hiring and promotion, including 15 percent who described it as a major or moderate factor.
Primary care physicians illustrate the pattern. They reported using AI more often than surgeons, 66 percent versus 58 percent, and were also more likely to say that staying current would give them an earnings advantage, 67 percent versus 55 percent.
Doximity also reported that 78 percent of physicians said AI had either increased their sense of job security or left it unchanged. That detail matters. The dominant mood in this dataset is not fear of disappearance. It is competition over who captures the time saved.
Is AI replacing doctors, or changing how doctors work?
The quotable answer from the 2026 data is this: 72 percent of physicians said AI is not changing the physician role in their specialty, but 50 percent said it is changing how work is done.
That distinction protects you from two poor decisions. The first poor decision is dismissal. If half of a large physician sample says the work itself is changing, ignoring the tools that change it is a career risk. The second poor decision is panic learning, where a doctor collects certificates without changing any real workflow. Hiring teams do not need another badge. They need evidence that you can safely get a clinic letter done faster, find the relevant guidance sooner, clean a messy audit dataset, or build a simple tracker that a department actually uses.
When AI enabled physicians to complete more clinical work in the same amount of time, 44 percent in the Doximity survey said physicians should benefit financially. That is not a settled pay model. It is a signal that productivity gains are becoming part of contract and promotion conversations. A doctor who can explain, in plain language, how a tool saved time without harming safety will be easier to promote than a doctor who only complains about the tool.
Why does AI fluency matter more for international medical graduates?
An international medical graduate rarely gets the easiest rota, the warmest introduction or the most forgiving first impression. You often have to prove competence twice, first on paper and then in the room. AI fluency helps because it creates portable proof.
A local graduate may rely on an old boys network, a familiar deanery or a supervisor who already knows their medical school. An IMG often relies on documents. A clean audit, a well structured quality improvement project, a teaching series with feedback, a portfolio that reads clearly. AI and data tools make those documents better and faster to produce, provided you use them with discipline.
Consider three ordinary examples.
A doctor who can take an exported list of clinic attendances, clean it in a spreadsheet, check for duplicates, and produce a simple run chart has something concrete for a portfolio. A doctor who can use an AI assistant to draft a first pass of a patient information leaflet, then edit it against current guidance and record the version control, has saved senior time. A doctor who can search literature, summarise what a paper actually did, and spot where the methods do not match the headline, is useful in a journal club and in a guideline meeting.
None of that requires you to become a programmer. It requires fluency, the ability to choose the right tool, check its output, and explain its limits. For an IMG, that fluency travels. Licensing exams differ by country. Good documentation, audit discipline and clear writing travel everywhere.
There is also a quieter advantage. Many IMGs work in posts with heavy service demand and little formal teaching. A careful AI workflow can turn a busy job into learning material. You can turn anonymised cases into reflection prompts, turn guideline updates into short quizzes for your team, and turn your own error log into a pattern you can fix. The tool does not do the learning for you. It removes the friction that stops busy doctors from learning at all.
What does AI fluency actually mean for a working doctor?
AI fluency is not prompt tricks. It is a small set of repeatable habits.
The first habit is documentation discipline. Use approved tools only where your organisation permits them, never enter identifiable patient data into a public system, and always read the final letter as if you wrote every word yourself, because professionally you did.
The second habit is source checking. An AI summary is a starting point, not a reference. Open the primary paper, the guideline or the report. Check the date, the population and the denominator. In the Doximity report, for example, the compensation data came from more than 23,000 responses collected in 2025, while the AI attitudes came from a June 2026 survey. Mixing those two samples would produce a confident error.
The third habit is simple data handling. Most clinical improvement work fails at the spreadsheet stage, not the statistics stage. Learn to tidy a table, define each column, count missing values, and make one honest chart. If you can do that, you can contribute to audits, service evaluations and rota planning in any department.
The fourth habit is writing clearly with help, then without shortcuts. AI can propose a structure for a referral, a presentation or a business case. Your job is to keep the facts, remove the filler, and make the recommendation explicit. British hiring panels still reward plain English. A fluent user sounds clearer, not more automated.
The fifth habit is governance awareness. Know who approved the tool, where the data goes, how long it is kept, and what happens when the output is wrong. A doctor who asks those questions early is trusted with pilots. A doctor who pastes first and asks later is not.
What should a doctor learn first, in practical terms?
Start with the work you already do every week.
In week one, map your repeated tasks. Letters, summaries, searches, teaching slides, audit counts. Pick one low risk task and build a template you can reuse. Measure the time before and after. Keep a note of errors you caught during checking.
In the first month, learn the spreadsheet basics that most doctors were never taught. Filters, pivot tables, simple charts, and a data dictionary. Add one AI assisted literature routine: define the question, search the primary source, extract the methods in your own words, and write what would change your practice and what would not.
By three months, aim for one visible output. A completed audit cycle, a teaching session with feedback scores, a one page service summary for your consultant, or a cleaned dataset that someone else can reuse. Portfolio evidence beats a course list because a reviewer can see the standard of your work.
For IMGs in particular, keep the evidence in a form that crosses borders. Save the final PDF, the feedback, the before and after counts, and a short reflection that names what you checked and what you would do differently. That habit will help in Denmark, the United Kingdom, Australia, Germany or anywhere else you build a career, because assessors everywhere ask the same question: can this doctor improve a service safely?
What are the risks of using AI badly at work?
The risks are ordinary, which is why they are dangerous.
Hallucinated references are the best known problem. An AI tool can produce a citation that looks real and does not exist, or a summary that reverses a result. The defence is boring and effective: never cite what you have not opened.
Confidentiality is the second risk. Patient identifiable information belongs in approved clinical systems, not in a convenient chatbot. If you would not email it to a stranger, do not paste it into an unapproved tool.
Automation bias is the third risk. A fluent sounding answer can lower your guard, especially at the end of a long shift. Keep a checking step that is separate from the generating step. Read the output after a pause, check numbers against the source, and ask a colleague to review high stakes documents.
Skill fade is the fourth risk. If a tool writes every reflection and summarises every paper, your own judgement gets less practice. Use AI to create first drafts and practice material, not to replace the final act of thinking. The doctors who will benefit most are those who remain able to work well when the tool is unavailable.
There is also a workforce context that should keep expectations honest. Doximity found that 85 percent of physicians said the physician shortage had already affected their clinical practice, and 76 percent said systemic pressures had compromised the quality of care they could provide in the past year. Reported overwork or burnout linked to the shortage eased to 52 percent in 2026, down from 63 percent in 2025 and 67 percent in 2024, yet broader strain remained high. In a separate June 2026 poll of more than 600 physicians, 82 percent said they were overworked, 46 percent were considering early retirement, up from 34 percent the year before, and 44 percent said they were somewhat or very likely to leave clinical practice within two years.
AI will not fix that on its own. In an overstretched service, a time saving tool can either give a doctor breathing room or give a system an excuse to add more work. That is why the 44 percent figure on who should benefit financially matters. Doctors need language for this conversation early, before productivity expectations rise without agreement.
Doximity also reported that average physician compensation rose by 2 percent in 2025, down from 3.7 percent growth the year before, and that the gender pay gap held at 26 percent for a second year. Those figures are United States specific, so do not import them into a European pay discussion. The transferable lesson is narrower: when overall pay growth slows, small differentiators, such as being the doctor who can run the audit, lead the digital pilot or teach the team, carry more weight in local decisions.
How can an IMG turn AI skills into portfolio evidence?
Make the work visible, specific and checkable.
Choose a problem your department already cares about. Missed follow ups, clinic letter delays, incomplete audit data, repeated teaching gaps. Write the question in one sentence. Collect a baseline count before you change anything. Use your AI and data skills to clean the data or draft the intervention, then record what a human checked. Remeasure. Write a one page summary with the chart, the limitation and the next step.
That single cycle, done honestly, is worth more than a long list of webinars. It shows clinical awareness, data handling, writing, governance and follow through. It also gives you a calm answer in interviews. When a panel asks how you keep up to date, you can point to a finished piece of work rather than a general claim about enthusiasm.
Keep the tone modest. State what the tool did, what you verified, and where it failed. Reviewers trust doctors who name limits. They worry about doctors who promise transformation.
The quiet advantage, stated plainly
AI fluency is becoming a doctor’s quiet career advantage because the profession is rewarding a new combination: clinical judgement plus the ability to handle information at speed, safely and in public. Doximity’s 2026 data suggest many physicians already believe this will affect earnings, hiring and promotion. The 67 percent and 39 percent figures may move up or down in future surveys, yet the direction is hard to ignore.
For an IMG, the opportunity is larger, not smaller. You may not control which country opens a training door this year. You can control whether your next application contains a clean audit, a clear teaching record and a portfolio that shows you can learn a system quickly. Those are AI assisted, human checked skills. Build them now, in ordinary posts, with ordinary tools, and keep the receipts.
That is professional currency. It spends in any health system.
Frequently asked questions
Do doctors need AI skills to get hired in 2026?
Not everywhere, yet the direction is visible. In Doximity’s 2026 report, 39 percent of physicians said AI proficiency already factors into hiring and promotion decisions in their specialty, including 15 percent who called it a major or moderate factor. Treat AI skills as a differentiator that supports clinical evidence, not as a substitute for it.
Will AI increase a doctor’s pay?
Some doctors expect it will. In the same Doximity report, 23 percent expected AI to increase their total compensation within the next 12 months, and 67 percent believed physicians who stay current with AI tools will have a meaningful earnings advantage over colleagues who do not adopt them. Those are expectations from a United States survey, not a guarantee in any single job.
Is AI replacing doctors?
Most physicians in the 2026 Doximity survey did not think so. About 72 percent said AI is not changing the physician role in their specialty, while 50 percent said it is changing how work is done. The practical task is to learn the changing workflow while keeping clinical responsibility.
What AI skills should a doctor learn first?
Begin with source checked literature searching, safe documentation templates in approved systems, spreadsheet cleaning for audits, and clear writing with careful editing. Add governance knowledge early: approved tools, data handling, checking steps and limits.
How can an international medical graduate use AI skills as an advantage?
Use AI and data tools to produce portable, checkable portfolio evidence: a completed audit cycle, teaching with feedback, a one page service summary, and a short reflection that states what was verified. Assessors in different countries can judge finished work even when training pathways differ.
What is the biggest risk of using AI at work?
The most common risks are hallucinated references, confidentiality breaches from pasting identifiable data into unapproved tools, automation bias, and skill fade. The safeguards are to open primary sources, use only approved systems, keep a separate checking step, and retain your own final judgement.
Sources consulted during drafting: Doximity 2026 Physician Compensation Report, released 25 August 2026, at https://www.doximity.com/reports/physician-compensation-report/2026 and Doximity’s summary of the same report. Figures on compensation growth, AI use, earnings expectations, hiring, job security, shortage and burnout are from that report as published.