ChatGPT Is Getting More Personal

This week, the most useful AI story is about memory and control.

ChatGPT is getting better at using what it already knows about you. That can make answers more useful, but it also makes the settings matter more.

OpenAI updated ChatGPT so it can give more personalized answers and show where some of that personalization came from. In its release notes, OpenAI says ChatGPT can now use saved memories, past chats, custom instructions, and, for some paid users, connected files or Gmail.

The important part is not just that ChatGPT remembers more. It is that ChatGPT is starting to show "memory sources," which let you see some of the information it used to shape an answer. If something is outdated, irrelevant, or just plain wrong, you can correct it, delete it, or mark it as not useful.

Personalization cuts both ways. An assistant that remembers your writing style, ongoing projects, or trip plans can save time. But if it remembers the wrong thing or pulls from an app you forgot you connected, the answer can drift off course.

OpenAI also released GPT-5.5 Instant as ChatGPT's new default model. The company says it gives clearer answers and is better at using context when that helps. The practical lesson is simple: the more personal these tools get, the more readers should know where the personalization is coming from.

OpenAI introduced Trusted Contact in ChatGPT, an optional feature for adults with personal accounts. If a user chooses a trusted person in advance, ChatGPT may notify that person after automated systems and human reviewers detect a serious self-harm concern. This does not replace crisis support or professional care. The takeaway is that AI companies are starting to build more human handoff points into moments where a chatbot should not be the only support.

Google announced new link-focused updates for AI Mode and AI Overviews in Search. Google says AI search responses will show more links inside answers, previews of websites, suggestions for deeper reading, and more visible public discussion sources. AI search is more useful when it gives you a way to click through, compare sources, and decide whether the answer deserves trust.

The U.S. government's Center for AI Standards and Innovation announced agreements with Google DeepMind, Microsoft, and xAI for pre-release model testing. Powerful AI models are getting more outside testing before release, especially around national security risks.

ChatGPT for Excel and Google Sheets

OpenAI now has a spreadsheet sidebar for Excel and Google Sheets. It can help build trackers, explain formulas, clean up messy sheets, summarize tabs, and make updates using normal language.

This may be useful if spreadsheets are part of your work or home life, even in a basic way. Think budgets, planning sheets, trackers, or reports you inherited from someone else.

The catch is that spreadsheet mistakes can matter. Duplicate important files first, ask for a plan before big edits, and review formulas or changed cells before relying on the output.

Memory sources

Memory sources are the bits of context an AI tool says it used to personalize an answer. That might include saved memories, past chats, custom instructions, connected files, or connected apps. If an answer feels weirdly specific or wrong, memory sources can help you see what the tool was leaning on and clean up old context.

Open ChatGPT settings this week and look for memory or personalization controls.

See whether memory is on, whether any apps are connected, and whether you can review what ChatGPT remembers about you. You do not have to change everything. Just learn where the controls live. If you see old context, clean it up before relying on personalized answers.

Thanks for reading. If this helped you feel a little more clear on what matters in AI right now, it did its job.

Know someone trying to keep up with AI without the jargon? Forward this to them or send them the subscribe link: https://ai-this-week-for-regular-people.beehiiv.com/subscribe

Follow AI This Week for Regular People on X: https://x.com/AIRegularPeople

See you next week.

This newsletter is for general informational purposes only. AI tools can be wrong, incomplete, or poorly suited for high-stakes decisions. Verify important claims and use qualified judgment before relying on them for medical, legal, financial, or other important personal or professional matters.