Direct answer
Direct Answer - ** You change ChatGPT's version with the model picker** in the message composer: open a chat, click or tap the model name at the top of the composer, and pick a different version from the list that appears [1]. The list only shows versions your plan and workspace are entitled to, so a missing option usually means a tier limit, an admin restriction, or a staged rollout rather than a broken app [1]. If the picker itself is absent, you are either on a client that has not received the update or signed into a workspace that has disabled model choice.
How Do You Actually Switch Between ChatGPT Versions in the App and on the Web?
The switching gesture is the same everywhere: the picker is anchored to the model name shown in the composer, and selecting an entry changes the model for the current conversation only until you change it again [2]. On the web and desktop apps the control sits in the composer row beneath the chat; on iOS and Android it appears in the same composer area, and the list mirrors exactly what the signed-in account is entitled to [2]. Because the picker is account-aware, the same phone can show a different list after you switch from a personal login to a university login.
Two practical details trip people up. First, changing the version mid-conversation does not rewrite earlier turns — the new model continues from the existing context, so quality shifts can look inconsistent if you switch halfway through a long draft [2]. Second, legacy versions disappear from the list on a published retirement schedule, so an option you used last term may simply be gone rather than hidden [2]. If you need reproducible output for coursework, start a fresh chat after switching so the whole draft is generated under one version.
A quick checklist that resolves most cases:
- Open any chat and locate the model name in the composer.
- Click or tap it and read the full list before choosing — availability differs by plan.
- Confirm the change by checking that the composer label updated.
- Start a new chat if you want the entire document written under the new version [2].
If the label does not change after selection, reload the client; stale sessions are a common cause of a picker that appears to ignore input [2].
Why Is the Model Switcher Missing or Greyed Out on Some Accounts and Plans?
The most common reason is tier gating. Free accounts see a narrower set of versions, while the newest flagship models are typically reserved for paid subscriptions, so an option that exists in a tutorial may never render for a free login [3]. Paid tiers widen the list but do not guarantee every entry, because OpenAI ships models in stages and two accounts on the identical plan can briefly see different menus [3]. That is why a classmate's screenshot may not match your screen even when your billing status is the same.
The second reason is administrative. On Enterprise, Edu, and Team workspaces, an owner or admin can restrict which models members may select, which greys out or removes entries that would otherwise be available [3]. University-managed accounts are frequently locked this way, and no amount of reinstalling the app will restore a control the workspace has disabled. If you are on an institutional seat, ask your IT or workspace admin before assuming a bug.
Regional availability and client version round out the list. A picker can be present on the web while the mobile build lags behind, and some versions are rolled out per region before going global [3]. When the switcher is genuinely absent, sign out and back in, update the app, and try the web client — if the web shows the list and the app does not, the constraint is the client build rather than your account [3]. Documenting which of these three causes applies to you saves a lot of wasted troubleshooting.
Does the ChatGPT Version You Used Change Whether Turnitin Flags Your Writing as AI?
Not in the way most students assume. Turnitin's AI writing report estimates the proportion of qualifying text in a submission that is likely AI-generated, and it reports that estimate per segment rather than naming a source model [4]. The detector works from statistical patterns in the prose you actually submit — perplexity, burstiness, and related distributional features — not from metadata about which version drafted it [4]. A paragraph written with an older model and a paragraph written with the newest one are judged on their surface texture, not their provenance.
What does change with version is the texture itself. Newer models tend to produce smoother, more uniformly paced sentences, which is precisely the pattern detectors are tuned to notice, while heavily edited or genuinely messy drafting reads differently [4]. This is why two students using different versions can receive very different scores on comparable work, and why a version upgrade can quietly raise your risk without any change in how much you personally wrote. The version is a contributing factor, not a verdict.
One display detail matters for interpreting results. When Turnitin's confidence falls below its threshold, the report shows an asterisked percentage instead of an exact figure, signalling a low-confidence signal rather than proof of zero AI use [4]. Treat that as a prompt to review the flagged segments, not as a clean bill of health. If your draft was produced or heavily assisted by a model and you need the score down before submission, the version you picked is only half the story — the wording has to change too.
Knowing which version wrote your draft is useful context, but it will not tell you what an instructor's report will say. turnitin0 lets you see that before you submit: upload your .docx, .pdf, or .txt (English, 300–30,000 words, under 20 MB) and you get a Turnitin AI detection report plus a similarity report, matching what markers see in their LMS, typically in under 15 minutes. If the AI score comes back high, the paired AI humanizer rewrites the flagged passages for ChatGPT, Claude, or Gemini drafts while keeping your meaning, citations, headings, and .docx formatting intact — with a score promise of *% or lower, or even 0%, or a full refund.
※ Turnitin0.com - [AI Humanizer](https://www.turnitin0.com/guides/us/turnitin-ai-humanizer-fix-what-to-do-when-your-real-writing-gets-flagged-as-ai) Clearing All AI Flag of ChatGPT Text
FAQ
Can I change the ChatGPT version in the middle of a conversation?
Yes — the picker stays available inside an open chat, and the new version continues from the existing context [2]. For a clean, reproducible draft, start a fresh chat after switching so every paragraph is generated under one version.
Why can't I see the newest model even though I pay for a subscription?
Model releases are staged, so eligible accounts can receive an option at different times [3]. Workspace admin restrictions and regional rollout are the other two common causes; check your seat type before assuming a billing problem [3].
Does switching to a newer ChatGPT version make Turnitin flag my work more often?
It can raise the statistical signal, because newer models tend to produce smoother, more uniform prose that detectors are tuned to notice [4]. Turnitin evaluates the submitted text's patterns, not the model that produced it [4].
What does an asterisked AI score in a Turnitin report mean?
It means the detected share fell below Turnitin's confidence threshold, so the exact figure is suppressed and shown as *% [4]. Read it as a low-confidence signal and review the flagged segments rather than treating it as zero AI use.
Do I need to re-check my draft after changing versions?
Yes. Because the version changes the prose texture, a draft that passed under one model may score differently under another [4]. Re-running a check on the final wording is the only reliable way to know what your instructor will see.