AI editing without surrendering the draft: testing Strut's passage-level and version workflow
A practical assessment of Strut.so's passage-level AI editing, version recovery, and collaborative workflow for long-form writing, with a test protocol readers can run themselves.
When a writer accepts an AI suggestion, the original draft disappears. A single click replaces hours of thinking with a machine's rewrite, and the earlier version is often gone. This is the problem Strut.so claims to solve: an AI workspace that keeps every passage, every draft, and every source connected, so you can edit with AI without surrendering control. For Hong Kong professionals writing reports, proposals, or academic papers, the ability to trace changes and recover earlier work is essential. This article provides a protocol you can run to assess Strut's passage-level editing, version recovery, source traceability, and collaborative features.
What Strut promises about version control
Strut presents itself as a long-form writing workspace where notes, sources, outlines, and drafts live in one project. Its interface shows a version history icon and a collaboration icon alongside the draft. The company states that you can "build and revise the full draft with context-aware AI that follows your direction." The key claim is that AI operations happen one passage at a time, and the writer can copy, insert, or replace suggested text. This suggests that the original passage remains accessible until the writer explicitly chooses to replace it. However, the publicly available information does not specify how long version history is retained, whether it supports branching, or how many versions are stored. You would need to test this by opening a project, making several AI edits, and checking the version history panel.
How to test passage-level editing
To assess whether Strut truly protects your draft, create a test document of about 500 words. Use the AI assistant to "Strengthen this passage" on a single paragraph. Observe the options presented: the interface shows a result with Copy, Insert, and Replace buttons. The safest workflow is to use Copy or Insert, which leaves the original text unchanged. Only Replace overwrites the original. A thorough test would involve using Replace, then immediately checking version history to see if the earlier version is recoverable. Repeat this with five different passages. If the original text is retrievable after each replacement, the tool passes the basic version recovery test. If not, you may need to manually copy passages before editing.
Source traceability: can you track where AI suggestions come from?
Strut allows you to attach sources to a project. The interface shows a "Sources" section within the project material. When you ask the AI to expand or strengthen a passage, the tool can reference those sources. But the public information does not detail whether the AI cites specific sources in its suggestions or simply uses them as context. To test this, upload two conflicting sources—for example, one report stating that Hong Kong's AI adoption rate is 30% and another stating 45%. Ask the AI to write a paragraph on adoption trends. If the suggestion includes a citation or clearly favours one source, traceability is strong. If it blends information without attribution, you must manually verify facts. The U.S. Copyright Office's Part 2 report (January 2025) notes that AI-generated outputs may not be copyrightable if the human's creative control is insufficient. Keeping source links visible helps document your editorial decisions.
Version recovery and the limits of AI tokens
Strut uses a token system: free users receive 2,000 AI tokens daily, up to 80,000. Pro users have unlimited tokens. Each AI operation—strengthen, expand, rewrite—consumes tokens. If you run out of tokens, you cannot use AI features, but you can still edit manually. Version history is separate from tokens; it should work regardless of your token balance. However, the company does not state how long deleted drafts are retained. A practical test: create a draft, make ten AI edits, then try to revert to the first version. If the version history shows each step with timestamps, recovery is reliable. If only the latest version is kept, you should export drafts regularly. For Hong Kong users handling sensitive data, the PCPD's Model Personal Data Protection Framework recommends minimising data retention and ensuring deletion upon request. Strut's privacy policy should clarify data storage location and retention periods.
Collaborative editing: who sees what?
Strut includes a collaboration feature, indicated by an icon in the project view. The public information does not detail whether collaboration is real-time or asynchronous, whether comments are supported, or whether version history tracks individual contributors. To test, invite a colleague to your project. Have them edit a passage and leave a comment if the tool supports it. Then check version history to see if the editor's name appears. If the tool does not attribute changes to specific users, you should establish a manual workflow: assign one person to make AI edits and another to review. The NIST AI RMF Playbook suggests measuring AI system trustworthiness through transparency and accountability. In a collaborative setting, each AI suggestion should be reviewable before it becomes part of the shared draft.
Limitations and privacy considerations
Strut is a web-based tool. Any document you upload is processed on its servers. The company states it is "free to use" with token limits, but does not specify whether uploaded documents are used for training or how long they are stored. Before using Strut for confidential work—such as client reports or legal documents—review its privacy policy and terms of service. Consider whether the data is stored in Hong Kong, the U.S., or elsewhere. The U.S. Copyright Office's Part 3 pre-publication (May 2025) addresses training data issues; if Strut uses your content to improve its model, that may affect your intellectual property rights. For most users, the practical advice is to avoid uploading sensitive information until the company publishes a clear data handling policy.
Practical checklist for readers
- Test version history: Make five AI edits using Replace, then check if each earlier version is recoverable.
- Verify source traceability: Upload two conflicting sources and see if the AI cites them in suggestions.
- Check token usage: Note how many tokens each operation consumes and whether the daily limit meets your needs.
- Assess collaboration: Invite a colleague and confirm whether changes are attributed to specific users.
- Review privacy policy: Look for data retention, storage location, and training data clauses.
- Export regularly: Download your draft as a backup before making major AI edits.
Strut offers a promising approach to AI-assisted writing that respects the writer's original work. But the proof is in the testing. Run the protocol above before relying on it for a critical project. Only hands-on use will reveal whether the tool truly protects your draft or merely promises to.
Sources and further reading
- Strut AI document editor
- U.S. Copyright Office: Copyright and AI
- NIST: AI RMF Playbook
- Hong Kong PCPD: Model Personal Data Protection Framework
Product features and terms can change; check the provider's current information before using a service for workplace material.
Guest Editor contributing to selected features about working life and technology.