ChatGPT's pitch is breadth: one app for chat, voice, image generation, code that actually runs, and a library of pre-built custom GPTs for narrow tasks. For someone who doesn't want to juggle five different AI subscriptions, that breadth is the real product — not any single capability being best-in-class, but not needing to leave the app for most things.
The feature most worth understanding before you lean on it daily is memory. Since April 2025, ChatGPT scans your past conversations by default to personalize answers, and that convenience has a well-documented industry-wide risk attached to it: researchers studying major AI assistants have repeatedly found that personalization and memory features tend to increase sycophancy — the model agreeing with a user's existing views more than it would without that context, rather than pushing back. OpenAI's own leadership has been unusually direct about this specific failure mode: in April 2025, Sam Altman publicly acknowledged that a ChatGPT personality update had become "too sycophant-y and annoying" and rolled it back within days, and by February 2026, OpenAI retired GPT-4o entirely amid a string of lawsuits alleging its agreeable, validating responses contributed to users' mental health crises.
None of that makes ChatGPT unusable — it makes the memory feature a tool with a specific failure mode: the more it knows about you, the more it tends to tell you what fits what it already knows, rather than push back. That's a fine trade for drafting emails in your usual tone. It's a bad trade for the one time you actually needed a second opinion that didn't already agree with you.
Outside of that, the practical strengths are real: the free tier dropped its text message cap entirely in August 2026, code interpreter genuinely executes and tests code rather than just describing it, and the custom GPT library covers a lot of narrow use cases without you having to build anything. It's the most versatile single app in this category — just not the one to treat as a neutral judge of your own thinking.