Key answer
Frequency alone is not the issue; consider whether you retain reasons, accept real-world feedback and can stop seeking approval. Frequent tool use differs from requiring its agreement before deciding.
Full explanation
Fang et al. ran a four-week randomized experiment with 981 people comparing chat modes/topics. Their October 2025 revision reported associations between greater voluntary use, trust/social attraction and higher dependence or problematic use. Usage duration was not randomized and there was no no-AI control, so associations do not establish that more use causes dependence. Study v2, results/discussion/limitations
Observe concrete changes:
· Do you form reasons first or repeatedly ask until receiving “you are right”?
· When AI conflicts with reliable material, real outcomes or trusted people, what do you verify?
· Are repeated rephrasings seeking the same reassurance without new information/action?
· Is pausing difficult, or is use displacing valued work, rest or relationships?
These are habits to observe, not a diagnostic scale. For a low-risk decision, write your judgment, reasons and uncertainty first. After AI advice, record only new facts/tradeoffs changing your view and decide in your own words. Compare with real outcomes rather than comfort alone.
Help inspect reasons and evidence, identifying missed options and costs. I will state my judgment first; then you add information, and I explain my decision. Do not end discussion with agreement alone.
If use persistently disrupts life, adjust boundaries and discuss actual effects with trusted people; AI need not decide whether you are dependent. Observe clearer reasons, progress in real action and continued choice to use or disengage.