August 6, 2026
Watch anyone use ChatGPT or Claude for a few weeks and you'll see the same thing happen over and over. They open a new chat, and before they get to the actual question, they spend a paragraph explaining who they are, what they're working on, and how they like things done. Every single time. Then they close the tab, come back tomorrow, and type the whole paragraph again.
Most people don't realize there's a setting that stops this.
Both ChatGPT and Claude now offer some version of persistent memory. Turn it on and the assistant can carry facts forward between conversations instead of starting blank every time. Not your whole chat history, just the useful stuff it picks up along the way. Your job, your preferences, the project you're mid way through, the way you like things phrased.
The easiest way to think about it: a new conversation used to mean a new employee on day one who knows nothing about you. Memory turns that into an employee who's been there a year and already knows how you like your coffee.
Two reasons, and they're opposite problems.
Some people never turn it on because they don't know it exists. It's buried in a settings menu nobody opens after the first setup, so it just sits off by default and nobody notices.
Other people turn it on and then never help it. Memory isn't magic. It picks up whatever crosses its path, so if your conversations are all short and transactional, it doesn't have much to learn from. You have to occasionally just tell it something directly.
"Remember that I run a small landscaping business in Ohio and most of what I ask you for is customer emails and social posts."
That one sentence does more for your future conversations than weeks of the assistant guessing from scraps.
Not everything. Memory shines on the stuff that repeats.
Your voice and preferences. Tell it once that you write short sentences and hate corporate phrases like "circle back," and it stops needing that instruction every time you ask for a draft.
Standing context about your work. Your job title, your industry, the names of your main projects or clients. Things that would otherwise eat up the first three lines of every conversation.
Ongoing threads. If you're working through a decision over several days across several separate chats, memory means you don't have to reconstruct where you left off each time.
Don't lean on it for anything you actually need right the first time. Memory is built to pick up general patterns, not to reliably store one specific number or date you mentioned once in passing. If it matters, say it again in the conversation where it matters, the same way you'd repeat a phone number back to someone instead of assuming they wrote it down.
It also isn't a replacement for good context on a real task. If you're asking for something complex, still explain the specifics up front. Memory removes the repetitive throat clearing. It doesn't remove the need to actually explain what you want.
Open the settings in whichever assistant you use and look for something called Memory or Personalization. Turn it on. Then, the next time you're mid conversation, just tell it a couple of things about yourself the way you'd tell a new coworker on their first week.
That's the whole setup. The payoff shows up quietly, a few weeks in, the first time you open a new chat and realize you didn't have to explain yourself at all.