OpenAI has announced some improvements to ChatGPT, which are said to make the tool more reliable and easier to use. The biggest change is to search, which is now said to deliver more accurate and useful results while cutting down on hallucinations.
According to the AI giant, factuality has been improved, which means ChatGPT will give fewer made-up answers and better overall response quality. Search also does a better job of understanding when users are shopping. If you’re looking for products, ChatGPT is more likely to show relevant options, and if you’re not, it will keep results focused on information rather than ads or product suggestions. Answers are now formatted to be easier to scan quickly without losing the detail or quality that regular users rely on.
Apart from that, personalization settings have been streamlined. Instead of opening in a separate window, users can now adjust personalization directly under Settings > Personalization, including choosing a ChatGPT personality, adding custom instructions, and updating Memory.
On a similar note, OpenAI recently introduced GPT-5-Codex for developers, which is a new variant of GPT-5 optimized for coding tasks in Codex. It’s available wherever Codex is used: as the default for cloud tasks and code review, and as an option for local workflows via the Codex CLI and IDE extension. GPT-5-Codex is designed specifically for coding work, while GPT-5 remains the go-to for general tasks. The company notes that GPT-5-Codex isn’t yet supported directly in ChatGPT or the API.
