On Review, Draft with AI writes a first pass into the three answer boxes and the one committed action, built from the trades, notes and figures already in that period. It is on Essential and Pro, and it exists for one moment: you have just spent twenty minutes rating trades, the questions are staring at you, and you have nothing crisp to say yet.
It is a draft, not a verdict. Nothing is saved until you edit it, and pressing Draft never finishes a review — that is still a button you press yourself, and it is what your streak counts.
What it is allowed to say
The draft is written from your own record: the period’s numbers, and the notes, lessons, mistakes and emotion tags you wrote on the trades. It is told, in as many words, not to invent an intention, a feeling, a plan or a market condition that is not there. Where your record is silent, a good draft says so — “both losses were adds to a position already offside; I did not write down what I was thinking on either” — rather than filling the gap. That sentence is usually the most useful thing on the page, because it points at the note that never got written.
Every figure it quotes is one this app computed. It is not asked to do arithmetic, so a draft cannot quote a P&L that disagrees with your dashboard. If a period is large, the draft is written from the same prioritised list the review pass itself shows you, and it is told how many trades it did not see so it does not describe a month it only half read.
Read it before you keep it. It is a machine’s first pass at your reasoning, and the one thing worth avoiding is finishing a review made of words you did not write and did not check. Change what is wrong and delete what you would not have said.
Drafting ONE trade’s note
Every card in the review pass also has a small Draft button above its “What happened on this one” box. It writes that trade’s note from the trade’s own numbers and anything you have already recorded on it.
This one is allowed to guess, and that is deliberate. On a trade with no note, no rating and no tag, the only hard facts are the price, the size and the result — a draft limited to those would just read the row back to you. So it may say what the tradelooks like it was, and it is required to mark every such sentence as a guess (“I did not write down why, but going by the numbers this looks like…”) and to end with a Still to fill in: line naming what only you know. It may never contradict something you did record, and it never invents a price or a date.
A drafted note is labelled until you edit it. The note carries a violet notice saying it was drafted by AI and that parts of it are inferred — on the review card, in your journal, and on your phone. That label is stored with the trade, so it is still there a month later, which is exactly when a guessed sentence is easiest to mistake for something you remember writing. The moment you edit the note the label disappears, because editing it is what makes it yours.
Every note has the same three headings — What Happened, Why I Did It, and Action Item for Next Time. The draft writes into that shape, and a Template button puts the empty headings into any blank note (on the review card, in your journal, and on the trade form) if you would rather write it yourself — each heading shows a faint example of what belongs there, which disappears as soon as you start writing that section. The middle one is the point: a box that just asks “what happened” gets answered with what happened, and the reason — the only part worth re-reading in a month — never gets written.
It writes about the WHY, not the numbers. Your size, prices, dates and result are on the screen right above the box, so the draft is told not to repeat them. What it talks about instead is how the trade was managed: whether a stop and a target existed, whether the stop was honoured, whether anything says the trade was planned.
Set your stop, target and rating first for a much better draft. With them the note can say whether the trade was managed to a plan; without them it is guessing from the price alone. If any of the three are missing the button asks you before it spends an allowance — you can always choose Draft anyway. The quickest place to set a stop and a target is the chart button on the same card: it opens the trade’s chart with both fields underneath it.
A trade that is still open cannot be drafted — there is no outcome to write up yet. Drafting a note spends from the same allowance as the period review above.
Your allowance
Review drafts have their own pool, separate from AI actions and from support messages — using one never costs you a trade review or a chart analysis. Unusually, it is capped over three windows at once: per day, per week and per month. Whichever runs out first is the one that stops you, and the message says which.
The reason is that the second draft of the same week is not a better review — it is the same evidence phrased differently. The caps are sized so that following the ritual as designed (a weekly pass, plus a monthly one) never reaches them; they exist to bound re-rolling, not to ration the feature. Settings → Billing shows all three at once.
A failed draft costs nothing. The allowance is only spent once you actually receive one, so a timeout or an error leaves your count where it was.
When the button will not work
- The period has no closed trades. There is nothing to draft from, and this is refused before any allowance is touched.
- The review is already finished. Re-drafting would overwrite what you decided to keep.
- You have already written something. The button still works, but it asks first — drafting replaces all three answers and your action.
See also: the weekly review and the AI features and how they are metered.