Choosing an AI writing tool: what to check before you sign up
Check five things before you pay for any AI writing tool: what happens to your text (training and retention), which plan you would actually be on, how billing works, whether the tool fits the job you do every week, and whether you can leave with your work. Test it on a real draft first.
Cite
DigiCruncher, "Choosing an AI writing tool: what to check before you sign up", https://digicruncher.com/topic/choosing-an-ai-writing-tool-what-to-check-before-you-sign-up/, accessed 2026-10-24.- Nathan ColeStaffEditor, DigiCruncher ·
Why the checklist starts with your text
Most disappointment with writing assistants is not about quality. It comes from finding out later that a free or personal plan treats your text differently from a business plan, or that a setting you assumed was on was off. The vendors publish this in their help centers. It takes about ten minutes to read, and it is the part of the decision you cannot undo once a client document has gone in.
The same product can have several sets of rules. Consumer plans, team plans and API access are often governed by different terms, and a setting on one does not carry over to another. So read the page for the plan you would really use.
The five checks
- Training and retention. Does your text train the vendor's models by default, can you turn that off, and how long is it kept either way?
- Plan and account type. Is the plan personal, team or enterprise, and who controls the settings: you or an administrator?
- Billing model. Flat monthly, per seat, or usage based? Watch for message caps, file size limits and features that sit on a higher tier.
- Fit for the job. A tool that is good for brainstorming may be poor at tracking changes in a 40 page document. Test your own task.
- Exit. Can you export your chats and documents, and delete your account without a support ticket?
What three vendors say, as of this writing
Policies move, so treat this as a map and not a contract. Each row links the page I read.
Tool Training on chats Retention and review notes ChatGPT The vendor says ChatGPT improves by training on conversations unless you disable "Improve the model for everyone" under Settings, then Data Controls Temporary Chats are not used to improve models while they stay temporary, and a copy may be kept up to 30 days for safety Claude (consumer plans) Used for improvement if you allow it, if a chat is flagged for safety review, or if you opt in to a program Incognito chats are not used to improve Claude. Commercial products and the API have a separate policy Gemini Apps Depends on the Keep Activity setting. Temporary chats are not used to train models Default auto-delete is 18 months, adjustable. Chats reviewed by human reviewers are retained up to three years. Temporary chats are held 72 hours Two things stand out. First, the "off" switch is real but it is not the same as "never stored": the pages above describe copies kept for safety, review or feedback in at least some cases. Second, business plans exist because the consumer settings are not designed for confidential work. If you handle client material, start from the business plan's documentation and ask your administrator, not from a personal account.
A ten minute test you can repeat
- Pick a real, non-confidential piece of your own writing, around 500 words, with an awkward paragraph in it.
- Ask each tool you are considering for the same edit: tighten, keep my voice, do not add facts.
- Check what changed. Did it invent a claim, drop a qualifier, or flatten a distinctive phrase?
- Run the same prompt again the next day. Consistency matters more than one good result.
- Try to export the result and the history. If that is hard now, it will be harder when you want to leave.
Matching the tool to the job
- Short copy and emails: almost any assistant is fine, so privacy settings and price decide it.
- Long documents: look at context handling and whether edits are shown as changes you can accept or reject.
- Grammar and style only: a dedicated checker may serve you better than a general chatbot, and often needs less of your text.
- Regulated or client work: the decision is a data protection decision first and a writing decision second.
Mistakes I see most
- Judging a tool on its best output instead of its worst.
- Assuming the free tier and the paid tier share the same data terms.
- Pasting something confidential "just once" before reading the policy.
- Choosing by a ranking list, which tells you what someone else needed.
Questions on a specific tool and plan are welcome below. Name the plan, what you write and who sees it, and people can be specific.
Write something first.
Give people something to work with: at least 30 words on what happened and what you tried.
That is too long. Keep it under 6,000 characters.
Write the question as the title, 15 to 140 characters, no links.
Pick a category.
Add a name (2 to 40 characters, no links).
That email address does not look right.
That was quick. Read the thread, then try again.
The form expired. Reload the page and post again.
Something went wrong with the form. Reload and try again.
Please complete the check and post again.
Limit reached for now. Try again later.
This thread is closed to new replies.
This thread no longer accepts replies.
Something went wrong with the form. Reload and try again.
Post a reply or question first (name and email), then this browser can vote, edit and accept answers.
You cannot vote on your own post.
Only the author (within 30 days) or the forum team can do that.
That email already has an account. Use your sign-in link instead.
New accounts are paused for the moment. Try again later.
That was already posted.