Of every occupation the Bureau of Labor Statistics identifies as affected by artificial intelligence, this is the one it expects to grow fastest.
The entry requirements are the unusual part. BLS lists the typical entry-level education as a postsecondary nondegree award, which means a certificate rather than a degree. Work experience in a related occupation: none. On-the-job training: none. Median pay was $50,250 a year, or $24.16 an hour, in 2024.
A certificate-level credential, no prior experience required, 14,200 openings a year, growth well above the economy, and work that is genuinely performable from home. That combination is rare enough to be worth examining carefully, including the parts that make it harder than the summary suggests.
The main thing the summary hides is a timing problem rather than a difficulty problem. Almost everything on this site can be started this week: the guides on freelance services, reselling, content and crypto all describe activities where the first pound or dollar is weeks away at most. This one is months away, because a credential and an experience requirement sit between you and the first paid chart. In exchange, what waits at the end is closer to a floor than to a lottery ticket, which is the opposite trade from most of the higher-ceiling activities covered elsewhere. Knowing which of those two shapes you are looking for is the decision that matters, and it is worth making before spending anything.
It is worth pausing on the apparent contradiction, because it explains the whole opportunity.
BLS explicitly names this field as one where AI is expected to have productivity effects. Its projections overview notes that medical secretaries and administrative assistants are expected to see productivity gains in billing and claims management, and that AI-based tools are affecting medical coding work directly.
And yet the projection is 7.1 percent growth, more than twice the national rate, while graphic designers sit at 2.1 percent and paralegals at 0.2 percent under similar pressure.
The reason is that AI exposure and employment growth are separate forces that can point in opposite directions. Automation reduces the labour required per unit of work. Demand growth increases the number of units. In healthcare, the second is winning: an ageing population generates more encounters, more procedures, more claims and more records, faster than software is reducing the effort per record.
That is a genuinely different situation from a stagnant field absorbing a productivity shock. It also tells you exactly which part of the job to avoid. The work that grows is not the part software is taking.
Two related jobs get bundled under one heading, and they pay differently.
The distinction matters for anyone deciding what to learn. Straightforward coding of routine encounters is the most automatable part of this field. Handling ambiguity, querying a physician about incomplete documentation, working a complex denial, and auditing another coder's work are not, because they require judgment about what the record means and communication with the people who wrote it.
Three systems do most of the work, and knowing which is which clarifies what any training programme is selling you.
The practical implication for choosing a credential: an outpatient-focused credential and an inpatient-focused one are training you on different code sets for different settings. Deciding which environment you want to work in should come before buying a course, and a surprising number of people discover the distinction afterwards.
The shape of this job, and its resilience, both follow from how it was created.
Diagnosis classification began as a public health exercise: standardising causes of death and disease so that populations could be compared across places and time. The International Classification of Diseases descends directly from that work, which is why the code sets are maintained internationally and why they carry statistical purposes alongside billing ones.
The billing function was bolted on later. As health insurance grew, payers needed a standard way to know what they were paying for, and the classification already existed. That combination, a statistical taxonomy pressed into service as a payment mechanism, explains most of the field's peculiarities: the extraordinary specificity, the frequent updates, the gap between a clinically accurate code and a reimbursable one.
Two developments made it a distinct occupation. Standardised electronic claims meant coding had to be consistent enough to be machine-processed, which created demand for people trained specifically to do it rather than for clinicians doing it informally. And the transition to ICD-10 in the United States, which expanded the code set enormously, made informal coding impractical and professional certification close to mandatory.
The relevant lesson for anyone entering now is that this occupation exists because of a permanent structural mismatch: clinicians document in prose for clinical purposes, and payment systems require structured codes. Software has been narrowing that gap for years and has not closed it, because the gap is not technical. It is the difference between how a human describes a patient and how a payment system needs the world represented. As long as clinicians write notes for other clinicians, something has to translate.
Two bodies dominate and both publish their fees.
AAPC requires membership to sit its exams, which is a further annual cost on top.
Some administrative details from AHIMA that are easy to get caught by. The eligibility window is 120 calendar days from application approval, and the exam must be scheduled and taken within it. Extensions come in 30-day increments up to 90 days at $50 each. Candidates who fail the CCA, CCS, RHIA or RHIT must wait at least 30 days before a new application is approved, and a retake requires a new application and fee. Exams are delivered at Pearson VUE proctored test centres.
The genuinely low entry cost here is real: a few hundred dollars for the exam plus study materials, against a median salary of $50,250. Few fields with that outlook have a credential this cheap.
Here is the part that the "no experience required" line in the Quick Facts obscures, and it is the single most important thing to understand before starting.
Pass AAPC's CPC exam without qualifying experience and you are awarded the credential with an apprentice designation: CPC-A. Employers filter on it. Removing the "A" requires, per AAPC's own guidance, one of:
- Two years of coding employment, or
- One year of employment plus education, or
- Practicode plus 80 hours of education
Read the first two again. Both require the job you are trying to get. That is a closed loop, and it is where most people who pass the exam stall.
The third route exists precisely to break it. Practicode is a practical coding module that AAPC accepts in lieu of experience when combined with education, and it costs money on top of everything already paid. For anyone entering this field without a healthcare job to start from, budgeting for that route from the beginning is the difference between a credential that opens doors and a credential that sits on a CV.
AHIMA's CCA is positioned as an entry-level credential and does not carry an equivalent apprentice marker, which is a reason some people entering cold start there instead. The tradeoff is that CCS and CPC carry more weight with employers once you have experience.
The honest framing: the exam is the cheap part and the experience requirement is the expensive part, measured in time. Anyone selling a fast route into this field is quietly skipping this section.
What It Actually Pays
The BLS median of $50,250 a year, or $24.16 an hour, describes employed medical records specialists across all experience levels and settings. Several things move a specific person around that figure.
Credential and specialty. Inpatient coding, which uses a different code set and involves more complex records, generally pays more than routine outpatient coding. Risk adjustment coding and auditing sit higher again, because errors there have direct financial and compliance consequences.
Setting. Hospital systems, physician practices, billing companies and insurers all hire coders and pay differently, with insurers and large systems generally at the upper end.
Productivity arrangements. Some employers pay per chart or per encounter rather than hourly, which rewards speed and accuracy and can pay well above the median for a fast, accurate coder. It also transfers volume risk to you.
Denials and audit work. Recovering denied claims and auditing other coders' output are the highest-value activities in the field for the same reason they resist automation: they require judgment and they directly produce money.
For side income specifically, the realistic shapes are part-time remote contract coding for a billing company, overflow work for small practices, and per-chart arrangements. The constraint is that most employers want the apprentice designation gone and some experience behind you, so this is more often a career transition with a side-income phase in the middle than a weekend activity from a standing start.
Remote Work, and Its Actual Conditions
This is genuinely remote-capable work, and the conditions attached are stricter than in most remote jobs.
Protected health information governs everything. You handle patient records, which brings real obligations: a private workspace others cannot see, a locked screen, no personal devices where the employer forbids them, no printing, and often employer-supplied equipment with monitoring. Some employers require a dedicated room with a lockable door and will verify it.
Productivity is measured precisely. Charts per hour and accuracy rate are both tracked, and accuracy is typically expected above a defined threshold. This is measured work in a way that most desk jobs are not, and people who dislike being metered dislike this job.
Fully remote roles frequently want experience first. Many employers hire on-site or hybrid initially and allow remote once you have demonstrated accuracy. That interacts badly with the apprentice problem for anyone whose whole reason for choosing this field is working from home.
The equipment bar is low. A reliable computer, a solid connection, and a second monitor that pays for itself immediately when you are reading a record and coding side by side.
Schedules are often less flexible than remote implies. Many roles have set hours because they sit inside a revenue cycle with daily deadlines, and some require availability during the practice's business hours for physician queries. Genuine choose-your-own-hours arrangements exist, mostly in per-chart contract work, and they are the minority. Anyone choosing this specifically for schedule control rather than location control should confirm which they are being offered before accepting.
Breaking In, Given the Apprentice Problem
The route that works is shaped entirely by the experience requirement, so plan around it from the start rather than discovering it after the exam.
Decide the setting before buying training. Outpatient physician coding and inpatient hospital coding use different code sets and lead to different credentials. Buying a course before deciding is the most common wasted expense in this field.
Budget the whole path, not the exam. Exam fee, membership, study materials, current codebooks, and the experience-substitute route if you have no healthcare job. The exam at a few hundred dollars is the visible cost and the smallest one.
Take any adjacent healthcare job you can get. Front desk, registration, medical records, patient accounts, prior authorisation, a billing company's data entry. These count toward employment-based apprentice removal, teach you how a practice actually works, and put you where internal coding vacancies get filled first. This is unglamorous and it is the highest-percentage move available.
Use the experience-substitute route deliberately if that door is closed. AAPC's Practicode-plus-education option exists to break the loop for people with no healthcare employment. It costs money, and the alternative is waiting for a job that wants experience you cannot get.
Consider starting at the entry-level credential. AHIMA's CCA is positioned for entrants and carries no apprentice marker. It is a weaker signal than CCS or CPC once you are established, and it can be the difference between getting a first interview and not.
Learn denials and documentation queries early. These are the parts of the job that resist automation and produce visible money. A candidate who can talk about why a claim was denied and how they would work it is interviewing for a different job than one who can only code a clean chart.
Expect the first role to be on-site or hybrid. Many employers grant remote status after demonstrated accuracy. If working from home is the whole reason you chose this field, know that it is usually the second job rather than the first.
Against the Alternatives
Worth situating, since several remote options on this site compete for the same person.
Against general virtual assistant or admin work. Coding pays meaningfully better at the median, has a credential that transfers between employers, and has a demand curve driven by demographics rather than by discretionary business spend. It also takes months of study before earning anything, where admin work pays immediately. The trade is time upfront for a floor underneath you afterwards.
Against other certificate-level healthcare roles. Most clinical certificate roles require physical presence. Coding is one of the few healthcare paths that is genuinely performable from home, which is its main structural advantage inside the sector.
Against the AI-squeezed professions covered elsewhere here. Graphic designers are projected at 2.1 percent growth and paralegals at 0.2 percent, both under AI pressure, against 7.1 percent here. If you are choosing a field to enter rather than defending one you are already in, that difference is the most decision-relevant number available, and it comes from the same BLS table.
Against freelance skilled work generally. Freelancing has a higher ceiling and no floor. Coding has a modest ceiling and, once you are past the apprentice barrier, a reasonably reliable floor with employer-side demand. It suits someone who wants stability more than upside.
The honest positioning: this is a career entry with an unusually cheap credential and an unusually good demand outlook, wrapped around a real experience barrier. It is a poor fit for anyone who needs income within weeks.
Who This Suits
Direct, because the study time is a real investment.
It suits people who are genuinely good at precise, rule-based work and enjoy it. The job is reading carefully and applying a large rulebook accurately, all day, with your accuracy measured. That is satisfying to a particular temperament and grinding to everyone else.
It suits people who want remote work with a floor under it. Few remote paths combine a certificate-level entry cost with demographically driven demand and a published median of $50,250.
It suits people already working in healthcare in any capacity. They have the hardest part, which is the employment that removes the apprentice designation, and an internal route to coding vacancies.
It suits career changers with a runway of a few months. The study, the exam and the apprentice route take time before the first payment.
It does not suit anyone who needs income quickly. There is no version of this that pays in week three.
It does not suit anyone who dislikes being measured. Charts per hour and accuracy rate are both tracked and visible.
It does not suit anyone uncomfortable with compliance responsibility. Your codes determine what is billed to insurers and public programmes, and that carries weight.
Rookie Mistakes
Believing the "no experience required" line without reading further. BLS is describing the education requirement accurately. AAPC's apprentice designation is a separate barrier and it is the one that stalls people.
Paying for an expensive programme before understanding the credentials. The exams cost a few hundred dollars. Training programmes vary by an order of magnitude in price and quality, and some are sold on the promise of a career that the apprentice requirement quietly delays.
Letting the eligibility window lapse. AHIMA's is 120 days from approval, extensions cost $50 per 30 days up to 90, and a failed CCA, CCS, RHIA or RHIT means at least 30 days before a new application is approved plus another fee. Book the exam when you apply.
Specialising in the most automatable work. Routine outpatient coding of straightforward encounters is where the software is strongest. Documentation queries, denial management, risk adjustment and auditing are where judgment is required and where the field's value is moving.
Coding from memory instead of the book. Experienced coders develop instincts, and instincts drift as rules change. The habit that separates accurate coders from confident ones is checking the current guideline for anything not entirely routine.
Underestimating the update cycle. Code sets and payer rules change on a schedule, and coding from last year's rulebook produces denials. Continuing education is a permanent requirement rather than a formality.
Treating accuracy as secondary to speed. Both are measured, and errors have compliance consequences beyond a productivity score. Incorrect coding can constitute fraudulent billing even when unintentional.
Assuming any remote job in this field is contractor work. Many are employment with schedules and monitoring. Read what you are agreeing to.
Gotchas Worth Knowing
Compliance exposure is real. Coding determines what gets billed to insurers and government programmes. Upcoding, whether deliberate or through carelessness, is a serious matter and the coder is not insulated from it. A coder asked to code in a way they believe is unsupported by the documentation has a genuine problem, and knowing that in advance is better than discovering it under pressure.
You will be asked to query physicians, and it is a skill. Incomplete documentation is the normal case. Asking a clinician to clarify without leading them toward a particular answer is a regulated communication with rules attached, and doing it badly is both ineffective and non-compliant.
Certification maintenance costs money and time. Continuing education units are required annually to keep a credential active, plus membership. Budget for the ongoing cost rather than only the exam.
Credential recognition is national but employment is not always. State licensure is not typically required for coding itself, but some employers and payers impose their own requirements, and roles touching certain programmes have additional screening.
The work is genuinely repetitive. Charts, all day, at a measured pace, with accuracy thresholds. People who need variety struggle, and the ones who last tend to be those who enjoy pattern recognition and precision.
Offshore competition exists at the low end. Routine coding is outsourced at scale. This pushes domestic value toward the work requiring clinical judgment, direct physician communication, and accountability under domestic compliance regimes, which is another argument for skipping the routine end.
The Billing Side, and Why Denials Pay
Billing gets less attention than coding and contains the better side-income opportunity, for a reason that is easy to miss.
A denied claim is money a practice has already earned and not been paid. The service happened, the cost was incurred, and the revenue is sitting with an insurer behind a rejection code. Someone who recovers it is not performing an administrative task; they are producing cash that would otherwise be written off.
That changes the pricing conversation entirely. Coding is a cost centre from the practice's perspective, and cost centres are squeezed. Denial recovery is measurably revenue-generating, which means it can be sold on outcome rather than on hours, and small practices are frequently terrible at it because nobody has the time to chase.
The work itself is unglamorous detective work. Read the rejection, work out whether the problem is the code, the documentation, eligibility, timeliness, authorisation or the payer's own error. Then fix and resubmit, or appeal with supporting documentation. Each payer behaves differently and much of the expertise is knowing those differences.
Two structural reasons this suits side income better than coding does. It is naturally project-shaped: a practice with a backlog of aged denials has a defined body of work with a measurable result at the end. And small practices are the natural customers, which means the buyer is a practice manager or the physician-owner who can decide without a procurement process.
The caution is the same one that governs the whole field. Resubmitting a claim with a changed code requires the documentation to support the change. Adjusting codes to secure payment for something the record does not support is not aggressive billing, it is a compliance problem, and the person who made the change owns it.
Behind the Scenes: A Day of Coding
The reality is quieter and more exacting than the description suggests.
You open a work queue: a list of encounters waiting to be coded, assigned by the system rather than chosen by you. Each one is a clinical record written by someone under time pressure.
You read it properly. Not skim, read, because the difference between two codes is frequently a single qualifier buried in a sentence about something else. Then you assign the diagnosis codes, then the procedure codes, then the modifiers that describe circumstances affecting payment.
Roughly one record in some steady fraction is ambiguous. The documentation supports two readings, or omits something the code requires. Now you decide whether to code conservatively or to query the physician, and queries slow your productivity metric while protecting your accuracy metric. Learning where that line sits is much of the first year.
Meanwhile the counter runs. Charts completed, time per chart, and an accuracy figure from audits of a sample of your work. Both numbers are visible to your supervisor and usually to you.
Periodically an audit comes back with a disagreement. Sometimes you were wrong, sometimes the auditor was, and there is a process. This is one of the few feedback loops in remote work that is genuinely rigorous, and it is how competence compounds here.
The rhythm suits a particular temperament: quiet, precise, uninterrupted, measurable. People who want that describe it as the best job they have had. People who want variety and human contact find it long.
Spotting the Bad Training Programmes
This field has an unusually large industry selling training into it, and the quality range is wide. The exam fees are published by the certifying bodies, which makes overpriced programmes easy to identify once you know the reference points.
Check what is actually included. A programme quoting a large number should be clear about whether it covers the exam voucher, membership, current codebooks, and any experience-substitute module. A course price with all of those excluded is a smaller purchase than it appears, in a bad way.
Be sceptical of guaranteed placement. No training provider controls employer hiring, and the apprentice designation is set by the certifying body rather than by the school. A guarantee that glosses over the experience requirement is the clearest signal to walk away.
Look for pass rates, and for what they are pass rates of. A programme's own internal assessment tells you nothing. Certification exam pass rates for their students, stated plainly, tell you something.
Prefer programmes aligned to a named credential. "Medical coding certificate" from an unknown provider is not a credential. CPC, CCA, CCS and RHIT are. The question to ask is which specific exam this prepares you for.
Watch for the setting mismatch. A programme teaching outpatient physician coding will not prepare you for inpatient hospital coding, and vice versa. If the marketing does not distinguish them, the provider is either unclear or being unclear on purpose.
Community colleges are often the better value. Many run programmes built around these credentials at a fraction of private-provider pricing, sometimes with local employer relationships attached, which is precisely the thing that solves the experience barrier.
The reference points worth holding: the exam itself is $199 to $499 depending on body, membership status and attempt count. Anything priced in multiples of that should be delivering proportionally more, and should be able to say exactly what.
Keeping the Credential Once You Have It
The ongoing obligations are modest but they are real, and people are caught out by them.
Continuing education is annual and mandatory. Both bodies require a defined number of units each year to keep a credential active, and letting one lapse means dealing with reinstatement rather than simply renewing.
Membership is a recurring cost. Member pricing on exams and access to some resources depend on it, and AAPC requires membership to sit its exams at all.
Codebooks update on a cycle. Working from a superseded edition produces denials, and the annual update is not optional. AHIMA notes its CCS exam moves to the 2026 codebook list from 1 May 2026, which is the kind of date worth tracking if you are scheduling an exam around a transition.
Specialty credentials compound. Adding a risk adjustment, auditing or documentation integrity credential to a base coding credential moves you toward the parts of the field that are growing rather than automating, and each is a smaller purchase than the first.
The reason to treat this as a plan rather than an afterthought is that the value in this occupation is drifting upward, away from routine coding and toward audit, integrity and physician-facing work. Someone who adds one specialty credential every couple of years is moving with that drift. Someone who passes one exam and stops is standing where the software is arriving.
Where This Goes Next
Four arguments about direction, drawn from changes already in motion. They matter mainly for deciding which specialty and which credential to invest in.
Demand growth continues to outrun automation, for now. The 7.1 percent projection reflects an ageing population generating more encounters faster than software reduces effort per encounter. That is a demographic trend rather than a technology forecast, which makes it more reliable than most projections in this space.
The routine end hollows out anyway. Autonomous coding of straightforward encounters is where the tools are best and where offshore competition is strongest. Expect the entry-level rung to get harder even while total employment grows, which is an uncomfortable combination for newcomers and an argument for aiming past it deliberately.
Auditing and integrity work grows fastest. When software codes a chart, someone must verify it, and the consequences of systematic error are financial and regulatory. AHIMA already credentials documentation integrity and health data analysis specifically, which is a signal about where the profession expects value to sit.
Physician documentation improvement becomes the adjacent prize. Most coding difficulty originates in incomplete documentation. The people who can work with clinicians to fix it at source are solving the upstream problem, and that work is not codeable by a tool because it is a conversation.
Remote becomes standard rather than a perk. The work needs a record, a code set and a secure connection, and employers have already discovered that accuracy is measurable from anywhere. The friction is compliance and monitoring rather than capability, and both are solved problems. Expect the on-site-first convention to weaken, which mainly benefits people who are geographically far from hospital systems.
The credential stays cheap and the experience stays the gate. Nothing suggests exam prices will move much, and nothing suggests employers will stop wanting demonstrated accuracy. Anyone entering should plan around the apprentice requirement from day one rather than discovering it after passing.