For practising NHS doctors
Write and review with us
We are building a question bank and a condition reference for doctors sitting the applied knowledge test and PLAB, on the rule that nothing reaches a reader without a specialty-matched doctor signing it off. That rule is only worth anything if there are doctors, and this is the page that asks.
Road to NHS has not launched. No item here has been through clinical review, every reviewer in the current bank is a development fixture rather than a person, and we have no users, no pass rates and no testimonials. This page will not invent any.
The two roles
What the work actually is
Two roles, and most people do one of them. You can do both, but never on the same item — the reviewer is always a different doctor from the author, and the database enforces that rather than trusting anybody to remember.
Authoring
You choose a condition or presentation from the published content map and give us the guideline passage a question has to be true to — the document, the section, the URL. A language model drafts against that passage and nothing else. You then edit the draft into something you would be content to have your name on: the clinical vignette, the single best answer, why each of the other four is wrong, and the teaching point.
The draft is a starting point, not a thing to rubber-stamp. How much of it you changed is recorded, and it is one of the numbers we watch — a run of items nobody edited is evidence that a prompt has drifted, not that the model got good.
Reviewing
You get items in your own specialty, in a queue, with the automated checks already run and their findings laid out beside the item — contradictions in the stem, implausible distractors, numbers outside physiological range, references that do not resolve or are out of date. You are reading a question that has already had its mechanical problems flagged, not proof-reading raw text.
You approve, ask for changes, or reject. If two reviewers disagree, nothing is averaged and nothing is majority-voted: two doctors disagreeing about a clinical question is information, and resolving it by counting throws that information away. It goes to the specialty lead with both rationales intact.
Your time
How long an item takes
We do not know, and we are not going to guess at you. Nobody has reviewed an item on this platform yet, so any figure here would be a number we made up to make the ask sound smaller. Here is what we can tell you instead.
The unit of work is one item
A vignette of roughly 120 words, five options, why the key is right, a line on each distractor, a teaching point, and a reference that resolves. No batch you have to accept, no minimum to commit to.
You set your own queue depth
An empty queue is a normal state rather than a problem. Nothing is auto-assigned to you beyond what you have said you will take.
We will time the first cohort
Per item and per role, and we will tell you the real number. When there is a measured figure it appears here, with the sample it came from.
Review time is not a league table
We record how long a review took as a health metric on our process. It is not ranked and not shown to other reviewers. If items in an area are approved in ninety seconds, the problem is the pipeline feeding them, not the doctor.
The tooling
What the tooling does for you
The honest version of “we value your time” is a list of the things you do not have to do. All of the below is built and working.
The checks run before you see it
Contradiction, distractor plausibility, numeric and unit sanity, reference resolution and currency, house style, and an originality check. Their findings sit beside the item with the reasoning shown, so you are arbitrating rather than hunting.
References are structured, not typed
You cite a source and the next re-verification date is computed from that source’s own cadence. You do not type a date, and you cannot forget one.
The queue is keyboard-workable
A queue of twenty should be doable without touching the mouse. There is a shortcut sheet on every screen that has shortcuts.
Nothing you approve is edited underneath you
Published content is versioned rather than mutated. The version you signed off is the version that stays attached to your name, and a later change is a new version with a new sign-off.
Guideline drift comes back to you
A background job watches the sources we cite. When one moves or goes past its re-verification date it returns to the queue of the doctor who owns that area, with what changed.
You can say no, and it is recorded as a no
Reject is a first-class outcome with a rationale attached, not a stalled item. So is “this needs a different specialty”.
Credit
You are named, and the credit is yours
This mark sits before the byline of everything that has been reviewed, and it opens the record: who wrote it, who reviewed it, GMC numbers, what depth of review the automated triage assigned and why, the guideline cited, and when it was last checked. A reader can press it on any item.
This is what the record looks like today: no calibration, no name, no date. The first line of it would be yours.
The record you would be signing
You appear on the public editorial board with your grade, your specialty, the areas you own, and the last four digits of your GMC number — so that a reader can find you on the register themselves. We hold the full number encrypted and never publish it.
Appearing is your decision and it stays your decision. The board page renders only for contributors who have opted in, and opting out removes you. Nothing we can do satisfies that consent on your behalf, and there is no code path in which it is assumed.
The board is currently not published, because nobody’s paperwork is complete. The board page states the standard and names nobody until somebody has met it.
You are also credited as an author or reviewer on every individual item you touched, for as long as that version exists. That is a citable body of work, and it is one of the few reasons a busy doctor should say yes to this rather than to something else.
The paperwork
The paperwork, before you ask
Four things, and none of them is optional — not as policy, but because the database refuses to publish content from a contributor missing any of them. You can author and read without them; what you cannot do is sign anything off.
A contract
A written agreement covering what the work is, what it pays and how it ends. You can stop at any time; what you have already signed off stays signed off, because the audit trail cannot be rewritten.
An assignment of intellectual property
The items you write for us become ours to publish. This is the ordinary arrangement for commissioned educational content and it is worth reading rather than skimming — it is the clause that matters most to you.
A current conflict-of-interest declaration
Industry relationships, paid advisory roles, holdings, anything that could look like a reason to prefer one answer. It expires and has to be renewed; when it lapses your ability to approve stops that day, automatically, rather than at the next review of anything.
A credential check, repeated yearly
We verify your GMC registration and re-verify it within twelve months. A check older than that stops you approving until it is redone.
Your employer almost certainly has rules about outside work. NHS trusts and boards generally require declaration of paid work outside your contract, and some require prior permission; the rules differ by employer, by grade and by nation, and a training programme may have its own. Doctors in training also have obligations to their deanery. This is your declaration to make and we cannot make it for you — but we will put anything in writing that your employer asks to see, and we would rather lose a contributor at this stage than find out later.
We are not in a position to give you legal or employment advice, and nothing on this page is that. Read the contract before you sign it, and take advice on it if you want to — that is a reasonable thing to do and we will wait.
The limits
What we will not ask you to do
Reproduce anything you remember from a real exam
Every item is original and written to the published blueprint. We do not ingest competitor content in any format, from any source, including a file somebody emails us. If you have seen a question, do not write that question.
Approve outside your specialty
The match is checked in the database and there is no override in the interface. If an item reaches you and it is not yours, send it back — that is a normal outcome, not a failure.
Approve at volume to hit a number
There is no target, no leaderboard and no throughput bonus. Triage decides how much review an item needs; it never decides whether review happens, and there is no threshold at which something is waved through.
Put your name to something a machine wrote unchanged
How much of the draft was edited is recorded, and a run of unedited items is treated as a problem with our prompt rather than as productivity.
Start a conversation
Leave an address and we will write to you — a person, about the specialty you work in and what you would want out of it. This is not an application and it commits you to nothing: no contract goes anywhere near you until you have asked to see one.
If you would rather see what the reader sees first, the condition reference and the registration pathway are public, and the privacy page says exactly what happens to anything we hold.