Switching AI models normally just means opening the site's own model list. Docognito keeps that same list one click away, as a chip in its menu bar, reading it straight from Claude or ChatGPT. A model locked behind a paid plan shows greyed out and cannot be selected through either version of the page.
Where the switcher lives in the document view
Docognito places a model chip in the menu bar of the document page, sitting where a document editor would show something like a document status rather than an AI setting, which is exactly the point of it. Clicking it opens a dropdown styled to match the rest of the page, listing the same models the real site would show if you opened its own switcher directly.
This placement is deliberate. A document editor does not normally carry an AI setting anywhere in its own menu bar, so the chip is designed to read as a small piece of document metadata, a status indicator rather than a control panel, the same instinct that shaped every other piece of the disguise around it.
The name showing on the chip before you click it is whichever model is currently selected for your next message, the same information the real site's own trigger button carries. It changes the moment you pick something different, and it also reflects a model the site chose on its own for a new conversation before you have touched the switcher at all.
Why the list is read, not stored
Every time the dropdown opens, Docognito reads the live model list directly out of the real page rather than working from anything saved earlier. That matters because your available models are not fixed. A plan change, a newly released model, or an account limit can all change what shows up between one session and the next, and a stored list would eventually go stale and start recommending something you no longer have.
This also protects against a specific kind of stale state, an account that had access to a model yesterday and lost it today, through a plan change or a temporary limit, would otherwise keep showing that model as available long after it stopped being one. Reading fresh each time closes that gap completely.
Reading it fresh each time also means the current selection is always correct. The checkmark next to a model in the dropdown reflects whichever one the real site has active right now, not whichever one was active the last time you opened the menu.
Locked models stay locked
A greyed out model is never clicked on your behalf, by design
Selecting a model your plan does not include is what triggers the site's own upgrade prompt. Docognito shows a locked model greyed out and leaves it there. It will not click it for you, because doing so behind a disguised page would pop that upgrade prompt somewhere you might not see it land.
This is a deliberate limit rather than something still being built. The switcher's entire job is to mirror a choice you could already make on the real site, not to grant access to something your account does not have. A model appearing in the list at all, even greyed out, only means the site itself shows it as an option, not that it is available to you.
What changes when you pick a different model
Open the model chip
Click the chip in the menu bar to open the dropdown, which lists every model the real site currently offers your account.
Pick an unlocked model
Selecting one applies it the same way it would on the plain site, ready for your next message rather than the one already sent.
Send your next message as usual
The reply comes back using the model you just chose, with nothing else about typing or sending any different from before.
Nothing about the conversation itself restarts or reloads when you change models. Both sites apply the change forward from your next message, and the document view simply reflects whichever model is now active, the same way the real page would.
Claude and ChatGPT keep separate lists
Claude and ChatGPT each maintain their own separate model lineup, and the chip shows whichever set belongs to the site you currently have open. There is no shared or combined list across the two, since Docognito reads each site's own switcher rather than keeping any list of its own to merge them into. Moving between the two sites simply changes which set of names appears when you click the chip.
This also means a change on one site never leaks into the other. Selecting a specific Claude model has no bearing on which ChatGPT model is active the next time you switch tabs, and the reverse holds too, since each chip only ever reflects the site it currently sits on top of.
One click to the same models your plan already allows
Use the model chip in the menu bar the same way you would use the switcher on the plain site, since it reads the same live list and applies the same choice forward to your next message. A greyed out entry means your plan does not include it, and that stays true whether the disguise is on or off.
Use Docognito to keep the switcher one click away inside the document view, without changing anything about which models your account can actually use.