ChatGPT Desktop: What Chat and Work Mode Means for AI Answers

Model: ChatGPT · Desktop App

On 16 July 2026, OpenAI rebuilt the ChatGPT desktop app around a single choice — Chat for quick questions or Work to finish tasks end to end — with a ChatGPT-and-Codex switcher, unified Recents, Projects in the app and Work that syncs across devices. Here’s what making agentic Work a default surface changes for whether AI names your brand.

Source: ChatGPT release notes — ChatGPT desktop app experience updates (16 July 2026)

What OpenAI actually changed

On 16 July 2026, OpenAI updated its release notes to describe a redesigned ChatGPT desktop app on macOS and Windows, live for all plans. The headline is a clearer layout built around one decision: a global switcher moves you between ChatGPT and Codex, and inside ChatGPT you choose Chat for quick, conversational help or Work to complete tasks end to end.

The rest is organisation. Chat and Work conversations now appear together in unified Recents, where you can sort, filter and pin them. Existing Projects show up in the app, and you can start a Chat inside a Project or open a Work thread using Project context. Cloud Work conversations sync across web, mobile and desktop, while local conversations stay on your computer. Codex remains a separate view with its workflows unchanged, and Quick Chats stay available for fast exchanges.

Why the Chat and Work split matters

On the surface this is a layout update, not a model change. But defaults shape behaviour, and OpenAI has just made agentic Work a co-equal, always-visible choice next to conversational Chat. When the interface nudges more people from “ask a question” toward “complete a task,” a larger share of journeys run through a mode that browses, researches and acts rather than only replying.

That matters for visibility because the two modes surface brands differently. A Chat reply is a shortlist you read; a Work task is a job the model carries out — comparing options, drafting, planning — and the brand it selects becomes the one it acts on. The logic that decides which names make the cut is the same one we set out in how LLMs decide which brands to cite.

What agentic Work changes for answers

Work is designed to finish things end to end, which means it does not just name a brand — it uses one. When a model completes a task, being the option it chooses is worth far more than being one line in a list, and being left out is more costly. It is the same shift we flagged with ChatGPT Work and, in Perplexity’s world, with Perplexity Brain.

The ranking machinery underneath is still ChatGPT, so who gets picked tracks the model and its grounding — the reason a change like GPT-5.5 Instant Mini or a model-picker change can move which brands appear. The desktop reorganisation does not rewrite that logic; it routes more people into the mode where it plays out. Strong, current public content still does the work, as we argued in is your brand ready for AI search?

How Projects context shapes answers

Projects arriving in the desktop app — and Work threads that run using Project context — push answers further toward personalised, workspace-grounded output. When a task draws on a user’s own files and project history, the shortlist bends toward what that context implies, much like the account-linked behaviour we covered in ChatGPT memory sources and when ChatGPT began bypassing memory for fast answers.

For brands, that reinforces a theme: there is no single universal answer to track. What ChatGPT names inside a Project can differ from a clean Chat, so measuring one prompt once understates the picture. This is the multi-context discipline behind building a GEO strategy — testing many prompts, personas and contexts rather than assuming one answer holds everywhere.

Why cross-device Work sync matters

Cloud Work now syncs across web, mobile and desktop, so a task started on a phone can continue on a laptop. That makes Work persistent rather than a one-off — a thread a user returns to over hours or days. Persistent, resumable tasks mean the recommendation moment is not a single query but an ongoing process, and your brand can be added or removed as the task evolves and the model gathers more context.

It also multiplies surfaces from one login: the same Work thread appears wherever the user is, echoing how quickly new answer surfaces have arrived — from ChatGPT Sites to Claude in Microsoft 365 Copilot. Each device is another place the same task, and the same brand choice, can play out.

How to check your presence in Work

Because Work now sits one click from Chat, it is easy to test only the conversational side. Here is a short, practical way to see how ChatGPT treats your brand once Work is a default choice — today is 19 July 2026:

1. Update the desktop app. On macOS or Windows, confirm you see the ChatGPT-and-Codex switcher and the Chat and Work choice.
2. Run the same prompt both ways. Ask your category’s buying question in Chat, then in Work, and record which brands each returns.
3. Give Work a real task. Ask it to compare options or plan a purchase end to end, and note the brand it acts on, not just mentions.
4. Test with Project context. Run the task inside a Project so it uses your files, and see how the shortlist shifts.
5. Check across devices. Start a Work thread on one device and continue on another to confirm the answer stays consistent.
6. Repeat over time. Defaults and models change, so one run is a snapshot; the value is in the trend.

What the desktop reorg means for brands

New answer surfaces have arrived in a steady stream — ChatGPT Work, ChatGPT Sites, ChatGPT on WhatsApp, Perplexity Brain and Claude Sonnet 5. The desktop update is different in kind: it does not add a surface so much as re-rank the ones already there, promoting agentic Work to a default. Interface defaults are quiet but powerful — they decide which mode most people use without anyone choosing it deliberately.

The practical response is the groundwork we keep returning to: clear, current content a model can read and reuse, so that when ChatGPT completes a Work task in your category, your brand is a natural pick. If your product could be reached commercially, our overview of ChatGPT advertising covers where paid placement is heading, and regional shifts like ChatGPT Ads in the UK show how fast the surfaces multiply.

How reconnAI tracks agentic surfaces

reconnAI monitors how the leading AI models — ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — answer questions across regions, and re-baselines that tracking whenever a platform changes where or how its answers are generated. A desktop redesign that makes agentic Work a default choice is exactly that kind of shift: the mode people use is changing, even when the underlying model is not.

If you want to see how ChatGPT represents your brand in both Chat and Work now the desktop app puts them side by side, get in touch with our team or explore how AI visibility tracking works.

ChatGPT desktop update: your questions

What changed in the ChatGPT desktop app? On 16 July 2026 OpenAI redesigned the macOS and Windows app around a global ChatGPT-and-Codex switcher and a Chat-or-Work choice, with unified Recents, Projects in the app, and Work that syncs across devices. It is live for all plans.

What is the difference between Chat and Work? Chat handles quick, conversational questions; Work completes tasks end to end, browsing and acting rather than only replying. Work is now a default choice sitting next to Chat.

Does this change what model ChatGPT uses? No. The update is an interface change. The model and its grounding still decide which brands get named, so the ranking logic is unchanged — only the mode more people land in has shifted.

Does it affect how my brand appears in AI answers? It can. As more journeys run through agentic Work, being the option the model acts on matters more than being one line in a list. Running your own prompts in both Chat and Work is the way to tell.

About reconnAI

reconnAI tracks how the major AI models represent topics and sources across ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — across multiple regions. We monitor how those models answer and how they change over time, so you can stay ahead of shifts in the AI landscape.

Last updated: 19 July 2026.