Review Mode
Review Mode turns a completed report into something you can sign off on. It renders the whole analysis as one continuous, editable document — like a word processor — where you turn Evalify's work-product into your own filed opinion: editing the prose and the structured findings in place, recording your professional judgment right where it belongs.
Every change you make is stored as a durable overlay that never alters the underlying analysis. Your edits save as you go, survive a re-run, and appear identically in the report, the shared link, and every export. It's built for the way an attorney actually reviews — at the level of a single claim element, a single flagged disagreement, a single sentence.
Review Mode is available on Counsel Grade analyses. What you produce is Counsel-Grade work-product — your judgment, durably recorded.
Turning it on
Open a Counsel Grade report and switch to the document view from the report's title bar. The document has two states, toggled with one button:
- Document preview — read-only, exactly as it will read and export.
- Editing document — the editable surface. Everything you can act on is marked in one consistent colour, so the review layer is easy to spot.
Leave editing and the controls disappear; the document underneath never changes. There's nothing to submit — your edits save and apply as you make them, and the Evalify Score updates on its own.
Editing the document
In editing mode the analysis reads top-to-bottom the way you'd read an opinion — executive summary, technical aspects, claim charts, products, patents, landscape, classification — and you edit it in place.
Write freely. Rewrite the FTO opinion, the section introductions, and each patent's claim-chart summary as your own prose. A compact formatting toolbar gives you bold, italic, underline, and links — your styling is preserved through saves, re-runs, and into the exported PDF.
Resolve a quality cross-check. Every claim chart is checked by a second, independent AI read. Where that second read disagrees with the main one, the report flags it as "needs your judgment." For each flag you agree with the lead read or agree with the cross-check (and can add a note). Your pick sets that element's verdict to the read you chose. If that flips a patent's overall conclusion, the score re-evaluates on its own.
Alter a verdict. On any row of the element-by-element claim chart, set the verdict yourself — Present, Absent, Partial Overlap, or Possible Doctrine of Equivalents. The row shows your call, marked as refined, alongside the AI's original read; reset it any time and the AI's read is restored. Your read is authoritative: if your changes flip a patent's overall conclusion — for example, marking every limitation present — that patent becomes blocking and the analysis re-scores to reflect your judgment.
Leave notes and edit mappings. Attach a note to any claim element or product mapping, or edit the mapping text itself — capturing the thinking you'd otherwise lose between sessions.
Exclude what's out of scope. From the claim chart or the aspect view, exclude a patent that doesn't belong. In Relevant Products and Adjacent Possibilities, remove items that aren't relevant — they're dimmed and struck through so you can see what you set aside, and you can undo it.
Add or edit aspects. Add, edit, or remove a technical aspect, or add a patent by number. These change what's searched, so they queue into a single re-run when you're ready — reviewing a dozen items never triggers a dozen re-analyses.
A refinement that doesn't change what's searched — re-including a screened-out patent, a verdict re-score — shows a compact "Updating your analysis" progress rather than the full patent-search screen: your existing search and charts are preserved, so it's much faster and your work is never re-run from scratch. A search-changing edit (a new or rewritten aspect) runs the fuller re-analysis.
Within the review layer, the action is legible by colour: modify edits, add actions, and remove/exclude actions each carry their own hue, so you can tell them apart at a glance.
What moves the score, and what doesn't
Most of what you record is a professional overlay — it refines and annotates the report without touching the score: resolving a cross-check, adding notes, editing mappings, refining prose, altering a verdict in a way that doesn't change a patent's overall conclusion, and excluding individual products or adjacent opportunities. These stand alongside the AI's work and are preserved whenever a re-run happens — a re-score never discards what you set aside.
Two kinds of change are substantial and update the score:
- Excluding (or re-including) a patent, or excluding an aspect — it changes what's in scope. The score re-derives on the next view.
- Altering verdicts enough to flip a patent's overall read — e.g. marking all limitations present turns a cleared patent into a blocking one.
Both are free (no credit, uncapped) and authoritative — applying keeps your read and never re-runs the patent search or re-charts over you. Only the score, the clearance verdict, and the written FTO opinion are re-derived from your edits, so it's fast, and the opinion is rewritten to match your conclusion. When a substantial edit changes the conclusion, Evalify regenerates the sentences that narrate it from your edited charts — never by re-searching — so the write-up can never contradict the verdicts beneath it. Adding an aspect or patent, which does change the search, queues into a single batched re-run.
If you rewrite the opinion and a later verdict change moves the conclusion out from under your words, Evalify flags the opinion with a "may be out of date" badge and offers a one-click revert to the regenerated text — so your prose is never silently left contradicting the findings.
A conclusion-changing edit — altering a verdict, excluding or re-including an aspect or a patent, or marking a patent as your own — likewise flags the Video Brief "out of date," with a one-click Regenerate. Regeneration is manual, since each render has a real cost, and the prompt is owner-only. It clears when you regenerate — or when you undo back to the exact state the current video already reflects, so a reversed change never leaves a stale flag.
When you clear patents yourself, the report credits your review: the Clearance Confidence panel shows how many patents were confirmed by a reviewer, and once you've confirmed that every charted patent is clear, the confidence reads High — your sign-off counts.
Undo and redo — the whole session
Review Mode keeps a full undo/redo history over everything you've done — verdicts, exclusions, notes, mappings, and prose — as one coherent session, not just your last keystroke. Use the undo/redo buttons in the document bar, or Ctrl/⌘ + Z to undo and Ctrl/⌘ + Shift + Z (or Ctrl/⌘ + Y) to redo. Undoing a substantial edit re-derives the score automatically. (While you're typing in a text field, the shortcuts do ordinary text undo, so they never fight the editor.)
Ask the assistant about your review
While Review Mode is on, the in-report AI assistant knows what you've done. Ask it to summarize your review, tell you what's still left to look at, or pull your decisions and notes into a short write-up — useful for handing a matter off. The assistant only reports on your review; it never makes the calls for you.
A quick tour
- Open a Counsel Grade report and switch to the document view, then Edit document.
- In the Detailed Claim Chart, open a patent. Where you see "needs your judgment," record whether you side with the lead read or the cross-check, and add your reasoning.
- On any claim row, refine the verdict if you disagree, and add notes where useful.
- Rewrite the FTO opinion and any section intros in your own words, with formatting.
- Exclude anything out of scope — patents, products, or opportunities — and apply queued changes when you've finished.
- Ask the assistant to summarize where things stand, then export the finished document.
See also: Filtering & Focus for narrowing the report to what you're reviewing, and Exporting & Sharing for turning your reviewed analysis into a client-ready document.