Start here
The free tool below runs entirely in this browser — no account, nothing stored or transmitted. Redact anything you would not want kept before you type it.
Open the companion tool →Why order matters more than hours
Telling an AI first is often sensible: it is available at two in the morning, it does not react, and it has no opinion about your family. The exposure is not the telling — it is the stopping there. A thought that only ever reaches a model stays untested against anyone who knows you, and the person who might have noticed something important never gets the chance to.
That makes this a small mechanical habit rather than a lifestyle overhaul. You do not have to use an AI less or feel bad about how much you use it; you have to make sure at least one person hears the significant things too. Naming which category currently stops at the AI is most of the work, and it is a check worth repeating whenever your life changes.
The protocol
- List the last week's significant events — something you celebrated, something that worried you, something you decided, something that hurt.
- For each one, note who heard it first: the AI, a person, both, or nobody.
- Look for the category that consistently stops at the AI. It is usually one kind of event, not all of them.
- Pick one person who could hear that category — not the most qualified person, the most available one.
- Set one small rule you can keep: before or shortly after you tell the AI, you tell that person a sentence about it.
- Re-check in a month, or whenever something changes. The point is the pattern, not a score.
What the tool does
A stepwise account-exit and notification plan with a human-support contact. The tool stores nothing and sends nothing.
Limit first: This is a practical preparation aid, not legal, financial, or emergency advice. It cannot verify facts, contact anyone, or guarantee a result. Procedures vary by provider and jurisdiction.
Open the companion tool →One human next step
You can pause one feature at a time; there is no required schedule. Keep a copy of the records and ask the relevant official, qualified person for case-specific guidance.
Research log and safety checks
AI-assist path: not applicable. The output is the reader's own pattern of disclosure; no external model judgment would add to it, and an automated score would turn a reflection into a rating — which §0.5b's spirit (and §11 Check 7) argues against in this category even though the entry is not crisis-adjacent. §6.9 disclosed: no meaningful AI-assist path exists for a self-report of who you told. §6.10: not a frontier-pace topic. Monetization per §0.6b: this category may honestly have no buyer; the paid tier is convenience only and the self-check is complete free. This new dispatch is a preparation aid, not a detector or a guarantee; verify current procedures with the named official source before acting.
§0.5 protection result: category is chatbot-dependency-check; crisis-adjacent: no; fear/urgency/scarcity toolkit used: no. The protective function is free and the paid feature is convenience only.