AI news and explainers at 7 AM weekdays, plus a Sunday weekly at 8Get it in your inbox

METAL LAB

ChatGPT's Visualize turns meeting notes into interactive screens

OpenAI demoed a skill that turns long documents into interactive screens you can share as images or websites

이미지: YouTube 영상 갈무리

Summary

  • On August 24, 2026, OpenAI released a YouTube video titled "Make information visual with ChatGPT," demonstrating how to use the Visualize skill in ChatGPT and Codex
  • Pasting in a long meeting transcript turns it into an interface that shows decisions, timelines, and open items at a glance, and you can ask for it to be reformatted, such as into a calendar view
  • The finished screen can be exported as an image or published as a website, letting you share it with your team via a link
Make information visual with ChatGPT
공개 영상
Make information visual with ChatGPT
공개일
2026년 8월 24일
채널
오픈AI 공식 유튜브
시연자
Katia Gil Guzman
적용 제품
챗GPT, 코덱스(Codex)
핵심 기능
문서를 인터랙티브 시각 인터페이스로 변환
내보내기 방식
이미지 저장 또는 사이트 게시 후 링크 공유

Meeting notes become a screen with tables and buttons

Instead of reading a long meeting transcript start to finish, you can paste it directly into ChatGPT or Codex and ask it to organize the information into a single screen. In a video titled Make information visual with ChatGPT, released on August 24, 2026, OpenAI's Katia Gil Guzman showed how this "Visualize" skill works. A skill, in this context, refers to a set of instructions that teaches the AI a specific procedure ahead of time — and this one takes what would normally be a text-only answer and turns it into an interface with tables, timelines, and buttons.

A long meeting transcript first turns into a screen organized by decisions, schedules, and open items. That screen is then revised repeatedly through back-and-forth requests to redraw it in other formats, such as a calendar. Once the desired format is ready, it's shared with the team as an image or website link.

The example in the video is a meeting transcript from event planning. Feeding these notes into ChatGPT produces a single screen laying out what's been decided, how the delivery schedule shook out, and which items are still unresolved. It doesn't stop there — asking for something like "show this as a calendar" prompts ChatGPT to redraw the same information as a calendar view, and you can keep iterating with further requests like this.

A feature that builds on the ChatGPT work flow

OpenAI has been rolling out content in recent weeks that treats ChatGPT as a workplace tool. As covered in ChatGPT Work: internal demo shows everything from meeting briefs to strategy decks, on August 19 OpenAI showed how sales, marketing, and strategy teams turn meeting notes and research materials into organized documents. This new Visualize video, which also centers on processing meeting notes, fits into that same broader push.

How to try it

The workflow shown in the video happens entirely within the chat window — there's no separate menu or button to click. The skill activates simply through what you ask for.

  1. Paste a document with a lot of information, like meeting notes, into a ChatGPT or Codex conversation and ask it to organize the content into a screen.
  2. Once processed, an interface appears showing decisions made, ongoing schedules, and items still undecided, all in one view.
  3. Asking for a different format — for example, "show this as a calendar" — redraws the same content as a calendar view.
  4. You can also ask for clickable buttons that send prompts back into the chat, turning the screen into something interactive that you can navigate across tabs.
  5. Once you're happy with the result, you can export it as an image or publish it as a website and share the link with your team.

The video confirms this workflow works in both ChatGPT and Codex. It didn't mention anything about plan tiers or country-specific restrictions.

What you can actually do with it

Gil Guzman said she uses this feature often — both as a morning briefing to quickly get up to speed on a topic, and to organize the status of ongoing projects. For example, you can pull together emails and notes from multiple departments into one view before starting your day, or share a project's timeline and open items with teammates through a single link. For teams with frequent meetings, this skill could cut down the time spent reorganizing notes every time.

StepInputOutput
1Paste meeting notes + request organizationInterface showing decisions, schedule, open items
2Request "show as calendar"Reformatted as calendar view
3Request to shareSaved as image or shared as website link

Editor's take

OpenAI's recent direction suggests a shift in emphasis — from a chatbot that just spits out text answers to a tool that builds the output into an actual screen. Anthropic's Artifacts and Google Gemini's Canvas move in a similar direction, but what sets OpenAI apart here is attaching this same capability to Codex, a coding agent, not just ChatGPT. Compared to the days when you'd get a document summarized as plain text, the real difference with this skill is that you can ask for the same content to be redrawn repeatedly in different formats — a calendar, a screen with buttons, and so on.

The busier an organization's meeting schedule, the more time tends to get quietly eaten up organizing meeting notes and research materials. Instead of manually copying that material into tables or slide decks every time, it's worth trying to paste it into ChatGPT and just ask for the format you want. That said, it's too early to tell from this video alone whether the feature is polished enough to replace formal internal reporting documents — teams will need to run it a few times themselves and check the accuracy of the output first.

In the coming weeks, it seems likely we'll see more examples of this skill applied to code reviews or project dashboards on the Codex side. Given that ChatGPT Work and Visualize have both come out of the same team in quick succession, it's fair to read this as OpenAI moving toward making screen output — not text output — the default way ChatGPT responds.

Comments