Review screen
Each item shows why it is there, the source, and the current text beside the proposed one. Proposed text lands only when you accept it. A change a run already applied stays unless you revert it.
On this page
How do I get a review, and what is each one for?
Four things put a review on your document. All four open the same screen.
- A workflow’s findings. A Find step, the dialogue or CTA pass, the legal clause flag or the AI audit staged what it found. Accept, fix or dismiss each finding. The review waits under N reviews pending above the first block. See Building & running your own.
- Changes a run already made. After Smart Proofread, Vision Proofread, Translate + Proofread, a workflow’s proofread or rewrite step, or Regenerate, Review changes on the finished message, or the run’s row under the reviews pending list, opens every block it changed. Keep or revert each one; a proofread’s change shows the note it was made for. A proofread fix on a block someone edited by hand is not applied: it waits here for Accept or Keep current. See Smart Proofread.
- Literess’s edits. When her approval card holds several block edits on one document, Review in one screen shows them side by side. Accept, edit or reject each. If you already have a review of that document open, her edits join it, and her card in chat drops its review link until you leave the screen. Your accepts and rejects reach her before her next step. See What Literess can do.
- Open comments. Review open comments in the Comments panel lists every open comment on the document, anyone’s, AI flags included. Resolve each, or get a fix for it. See Comments & real-time editing.
How does the review screen work?
Each item is a card: why it is here (the flag’s reason or the comment), the source beside the current text, and the proposed text as a marked-up diff.
Click the proposed text to edit it in place.
Get a fix (or Get 3 options) asks a model for a suggestion, with an optional instruction. Try again asks for a different one.
| Item | What you decide |
|---|---|
| Fix (a find’s suggested fix) | Accept, Keep current, or Skip |
| Flag (an AI comment on a block) | Resolve, keep text, accept a fix you asked for, or Skip |
| Match (a block a review-only find matched) | Pick one of the options and Accept, Keep current, or Skip |
| UI label (a label the UI-label check caught) | Accept the restored label, Keep current, or Skip |
| Applied (a change a run already made) | Keep or Revert to previous |
| Literess edit | Accept her text (edited or not), Reject, or Skip |
| Comment (anyone’s open comment) | Resolve, keep text, accept a fix you asked for, or Skip |
Reverting a change a proofread made adds a reply under your name to the note it was made for: “Reverted this change and kept the previous wording.” The next proofread reads that reply and leaves the wording alone.
- Accept saves the text as a new version of the block and closes the comments behind the item.
- Skip leaves the item open for later.
- Undo reopens a card you accepted, kept or skipped. On an accepted card, it also puts back the text the block had before.
Which settings apply to the whole review?
Three settings sit at the top of the screen, and you can change them any time:
- Layout: List shows every card on one page; One by one shows one card at a time, with keys for each decision (Enter accept · R suggest · E edit · S skip · 1–3 pick · ←→ move).
- Suggestions: what Get a fix writes, One fix or 3 options to choose from.
- Model: which model writes suggestions, Fast, Standard or Pro. The cost shows next to the button before you click.
Can I decide many items at once?
Yes. The bar above the cards has a button for each thing you can do to everything still open:
| Button | What it does |
|---|---|
| Accept all (N) | Accepts the chosen text on every card that has one |
| Fix all (N) or Options for all (N) | Writes a suggestion onto every card without one, then leaves the review open so you can check them; shows the most it can cost first |
| Keep all (N) / Revert all (N) | Keeps or reverts every change a run already applied |
| Resolve remaining (N) | Resolves every open flag and comment that has no suggestion on it; the text stays as it is |
| Skip remaining (N) | Leaves the rest undecided and finishes the review; their flags stay open in Comments |
Each one asks for a second click (“Click again to confirm”), so a stray tap cannot close a long list.
Cards you already decided keep their decision.
Where do I find a review, and can I leave halfway?
A review never opens by itself. It waits in the editor above the first block, under N reviews pending, one row per language for a multi-language run.
The Workflows tab lists the same reviews under Reviews ready.
Open a review when you’re ready. The button at the top left leaves it at any time and takes you back to where you came from: Back to editor, Back to review when you opened this review from another one, or Back from any other page.
What you did there is kept, on any device: the suggestions you asked for, the options you picked, text you edited, what you skipped.
The row shows how far you got, for example “3 of 12 done”, and stays listed until you finish or dismiss it.
A Paused badge means the workflow stopped mid-way to wait for you. Its later steps run only after you approve.
Dismissing a paused review cancels those later steps. You’re asked to confirm first.
You pay only for work already done. A review that waits for you never holds your credits.
What can Literess do on the review screen?
Literess sees the review you have open. Ask her and she switches its layout, suggestions or model. Ask her to change a block and her version appears on the screen at once, whatever the review is about. On a card with options, it joins them as one more option, signed Literess and already picked. Otherwise it arrives as its own card marked New, under that block’s card or at the end.
A note at the top names the block, with a Show link. Her version waits for your Accept like any other card, even with Allow All on.
She can also pick one of a card’s options for you, add her own wording as a new option, or highlight a card or a button to show you where to look.
Ask her to press a card’s button, such as Accept or Skip, and she does it only with Allow All on. Otherwise she tells you which button to press.
How do I choose whether a step waits for my review?
Each review step in a workflow has a Review setting:
| Review setting | What happens |
|---|---|
| No review — just flag it | Findings land as comments plus a sidebar report, with nothing to review |
| Review when the run finishes | Later steps run anyway; the review waits for you afterwards |
| Pause and review before the next steps | The run stops and shows Paused until you decide |
How to review sets the layout the review opens in: List or One by one.
Two more settings on the step set what the review screen writes when you ask for a suggestion:
- Suggestions on review: One fix or Three options. Match the find mode gives options for a review-only find and one fix otherwise.
- Model for suggestions: Fast, Standard or Pro. Automatic uses Fast for options and Standard for fixes.
All three are starting points. The bar at the top of the review screen changes them for the review you have open.
What does reviewing cost?
Deciding is free: accepting, keeping, reverting, resolving, skipping and editing cost nothing. The run that produced the findings was billed when it ran.
Asking for something new costs credits, and the button shows the price before you click. A fix and three options cost the same: the block’s source words at the model’s rate.
| Model | One fix or 3 options |
|---|---|
| Fast | 1 credit per source word |
| Standard | 1 credit per source word |
| Pro | 3 credits per source word |
A fix costs nothing when the model finds nothing to change.
Why does one review open as a list and another one by one?
Each review opens with its own defaults: a review-only find opens one by one with three options, and flags and suggested fixes open as a list with one fix. The built-in templates write the options on Standard; a step you add yourself writes them on Fast unless you pick another model. The step’s review settings change those defaults, and the bar at the top of the screen changes them for the review you have open.
Can I make the rest of the workflow run on only the findings I picked?
Yes. Set that step’s Review to Pause and review before the next steps. The run stops there so you can pick which findings carry forward, and Apply & continue runs the remaining steps on just the ones you kept.
Why can’t I see the glossary and style-guide panels while I am reviewing?
The review screen shows only the review, with Literess beside it. Back to editor takes you to the panels, and the review keeps your place. A glossary or style guide you attach in the meantime takes effect on the next run.
Still stuck? Ask Literess in the app, or write to support@transept.ai.