Model: ChatGPT Deep Research · OpenAI
On 9 September 2026, OpenAI brought Deep Research to ChatGPT Work and Codex — the mode that browses the web, reads your files and writes up an editable, cited report. As one of ChatGPT’s most source-transparent surfaces reaches deeper into the working day, which brands it cites starts to matter a great deal more.
Source: OpenAI ChatGPT release notes — Deep Research in ChatGPT Work and Codex (9 September 2026)
- On 9 September 2026, OpenAI made Deep Research available in ChatGPT Work and Codex, not just standard chat.
- It researches across the web, your files and connected apps, then turns the findings into an editable document, presentation, spreadsheet or Site — with citations.
- You can steer it while it runs; start by typing @Deep Research in Work or asking for deep research explicitly.
- It is available to Plus, Pro, Business, Enterprise and Edu users with Work access on web, desktop, iOS and Android, using existing Work or Codex allowances.
- Deep Research shows the sources behind its answers, so where your brand appears in those cited reports is now a visibility question worth tracking.
What OpenAI actually launched
On 9 September 2026, OpenAI made deep research available inside ChatGPT Work and Codex, extending a mode that had lived mainly in standard chat. In OpenAI’s own description, it lets you “research complex questions across the web, your files, and supported connected apps, then turn the findings into an editable document with citations.” You can steer the research while it runs and ask for the output as a document, presentation, spreadsheet or Site, depending on the tools available to you.
Getting started is simple: type @Deep Research in Work, or explicitly ask for deep research. It is available to Plus, Pro, Business, Enterprise and Edu users with Work access, on web, desktop, iOS and Android, and it draws on your existing Work or Codex allowance rather than a separate pool — deep research limits in ordinary Chat are unchanged. Crucially, this is not a new model; it is a research workflow, powered by OpenAI’s current models such as GPT-6 Astra, being promoted into more of the surfaces where real work happens.
Why Deep Research matters for visibility
Most AI answers give you a conclusion and, sometimes, a link or two. Deep Research is different: it plans a question, opens and reads multiple pages, cross-checks what it finds and then hands you a written report with its citations attached. That makes it one of the most source-transparent surfaces in ChatGPT — you can see exactly which pages shaped the answer. When a mode like that moves from a niche feature into Work and Codex, far more high-stakes, decision-shaping outputs start being built on cited web sources.
The way any of these tools choose what to cite follows the logic we set out in how LLMs decide which brands to cite: they lean on the clearest, most consistent public evidence they can find. A research mode that reads several sources before it commits applies that logic more rigorously than a quick answer ever could — which is exactly why being present, and clearly evidenced, in the material it reads is worth your attention.
What a citation-first research mode changes
Because Deep Research compares sources rather than trusting the first one it meets, it rewards brands whose story holds up under scrutiny. If your key facts — who you serve, what you cost, where you operate — line up across your own site and third-party sources, a research run can cite you with confidence. If those facts conflict or are hard to find, a cross-checking model is more likely to set you aside in favour of a clearer competitor.
This raises the premium on being genuinely machine-readable. Clean structure, plain statements of who a product is for, and consistent details are what a planning-heavy workflow can lift and reuse, exactly as we describe in is your brand ready for AI search? It is the same groundwork that helped brands appear in Perplexity’s deep research — the pattern now repeats on the biggest answer surface.
What Deep Research in Work and Codex changes
OpenAI says Deep Research turns findings across the web and your files into an editable document with citations.
Where Deep Research finds its sources
Deep Research pulls from three places: the open web, the files you give it, and supported connected apps. The web is the part you can influence for a broad audience — it is where a research run discovers brands it was not already told about. Being findable and clearly described on the open web is therefore the difference between being considered and being invisible when someone runs a category question through the mode.
The output can be a document, presentation, spreadsheet or even a ChatGPT Site, and it lands inside ChatGPT Work — the agentic workspace we have tracked as it has grown. That matters because a cited report is not a fleeting chat reply; it becomes a document people share, revisit and act on. The sources it named travel with it, which is why the same evidence discipline that shapes how ChatGPT remembers sources applies here too.
Does appearing in a Deep Research report help you?
It cuts both ways, as deeper capability always does. A mode that reads and compares more thoroughly is better at recognising a genuinely strong fit — so if your public evidence clearly matches the question, a research run is more likely to surface and cite you. If your presence is thin, dated or contradictory, that same thoroughness passes over you with more confidence, because it has seen the alternatives.
There is also a quality filter at work. A report that cross-checks claims is less easily swayed by thin or manipulative pages — the sort that spam policies applied to AI responses are designed to catch. The reliable path is not a trick but substance: clear, corroborated evidence that a careful reader — human or model — would trust.
What this means for your content
The response is not to chase Deep Research specifically, but to make your fit unmistakable to any capable model. Spell out who your product is for, the problems it solves, the constraints it suits and the outcomes it delivers, in clean, structured content a model can read and reuse. That same groundwork serves you across Gemini, Claude and Perplexity alike, and it is the heart of a durable GEO strategy.
Consistency across the open web is what a research-first workflow rewards: the more your key facts line up across your site, your profiles and third-party sources, the more confidently a run can cite you without hedging. It helps, too, to earn a place in the trusted, third-party material these tools favour, much as Google leans on preferred sources in AI Overviews. If your category can be reached commercially, our overview of ChatGPT advertising covers where paid AI placement is heading — but the organic groundwork comes first.
How to check your Deep Research visibility
Because Deep Research now reaches into Work and Codex, a report your customers or competitors run could shape a real decision. Here is a short, practical way to see where you stand — today is 11 September 2026:
- Run a real research question. Use @Deep Research in Work to ask the buying or comparison questions your customers actually ask in your category.
- Read the citations, not just the answer. Open the report and note which pages it cited — and whether any of them are yours.
- See who it named. Record which brands appear in the write-up and how they are described.
- Check your evidence trail. For any claim the report makes about your category, find where that fact lives on the open web and confirm it points to you.
- Test more than once. A research run can vary, so repeat the question and compare which sources and brands recur.
- Track the trend. One run is a snapshot; the value is in watching how your presence in these cited reports moves over time.
How reconnAI tracks ChatGPT answers
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 how it works. Deep Research reaching Work and Codex is exactly that kind of moment: a source-transparent, citation-heavy mode is about to shape far more of the answers your customers rely on.
It joins a busy run of shifts we have tracked lately — from ChatGPT’s redesigned desktop app to bigger custom instructions. If you want to see how ChatGPT represents your brand when it researches a question in depth, get in touch with our team or explore how AI visibility tracking works.
Deep Research in ChatGPT: your questions answered
What is Deep Research in ChatGPT? It is a mode that researches a complex question across the web, your files and connected apps, then turns the findings into an editable document, presentation, spreadsheet or Site — with citations. As of 9 September 2026 it is available in ChatGPT Work and Codex, not only in standard chat.
Who can use it, and how do I start? It is available to Plus, Pro, Business, Enterprise and Edu users with Work access, on web, desktop, iOS and Android. Type @Deep Research in Work or explicitly ask for deep research; it uses your existing Work or Codex allowance.
Is Deep Research a new model? No. It is a research workflow that runs on OpenAI’s current models rather than a model launch in its own right. What is new is where it is available and how much of the working day it can now touch.
Does appearing in a Deep Research report boost my visibility? Being cited puts your brand in front of the person reading the report, and those citations travel with the document. It is earned by clear, consistent, corroborated evidence on the open web — not by a trick — and it is worth measuring across every model, not ChatGPT alone.
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: 11 September 2026.