Open in app

Sign in

Write

Sign in

Integrating AI into Windows (a low-tech approach with AutoHotKey)

David Gilbertson
8 min readJul 17, 2024

--

Wouldn’t it be nice if you could press a shortcut and ask ChatGPT about something you see on your screen? Or select a bit of text in any application, press a shortcut, and have ChatGPT explain what it means?

In this post I’ll show you how to set up these two interactions.

To be clear, what I describe below doesn’t allow you to achieve anything you couldn’t do before, it just makes it easier; it makes it feel like your favourite Chatbot is integrated right into the operating system, rather than having to copy/paste between windows.

These instructions are for Windows, Chrome and ChatGPT, but you can apply the same logic to get the job done with different tools.

Setup

AutoHotKey

First, install AutoHotKey (apologies if you don’t like doing that sort of thing, I too avoided it for a few decades).

This is a tool that allows you to write scripts that run in response to a keyboard shortcut.

Next, create a file anywhere (e.g. your documents folder) called shortcuts.ahk. Leave it empty for now.

This file will define the keyboard shortcuts and the scripts that run in response to them. To make the file ‘active’ you can double click it, but if you want these shortcuts always available, you’ll want the file to run on startup. So:

  • Copy the shortcuts.ahk file.
  • Press Win + R, type shell:startup and hit Enter. This will open your startup directory. Everything in here is executed when Windows starts.
  • Hold down shift and right click in a blank area (this gives you the hard-core right-click menu) and select Paste shortcut.
  • Now double-click the shortcut to test that it launches AutoHotKey — you should see an H icon show up in your taskbar. If you hover, it should say ‘shortcuts.ahk’.

If so, AHK is working, ready for us to define some shortcuts. But first…

Chatbot as a window

As we write the scripts below, we’ll need a way to bring up an AI chatbot window. If your preferred chatbot comes as a Windows app, great, if not we’ll ‘install’ the website.

Depending on which chatbot you use, you might see a little Install button on the right of the URL bar. For example, to install Claude there’s a button:

Otherwise you might need to click on the snowman buttons in the top right of the browser and go to Save and share > Install page as app…

When asked if you’d like to pin the app to your taskbar, click Yes.

Now you should have your chatbot installed so it behaves like a Windows app.

Note that if you drag the pinned icon into the first position in your taskbar (just to the right of the start button), you can open it with Windows + 1. This works for other numbers and positions too. This is one way to open the app with a shortcut.

If you don’t want to pin the app to your taskbar, or rely on its position, you can give it a shortcut like so:

  • Right click its icon in the taskbar.
  • In the right click menu, right click the app name (right click inception)
  • Click Properties
  • In the dialog that appears, define a shortcut in the Shortcut key box and click OK.

With that done, we’ll be able to open the chatbot window with a shortcut.

OK, let’s get to the shortcut scripts. You can edit the shortcuts.ahk file in any text editor. If you’ve got IntelliJ or VS Code, I’d suggest installing the AutoHotKey plugin to get syntax highlighting.

To avoid clashing with shortcuts in other apps, I like to use The Holy Trinity of ctrl + alt + shift, and preferably use a key close to these so I can reach them all with my left hand.

With that in mind, I’ll use ctrl + alt + shift and:

  • A to Ask ChatGPT about some text
  • S to Show ChatGPT part of the screen

Asking about selected text

Here’s the plan:

  • When you press the shortcut ctrl+alt+shift+a, AHK will run a little script.
  • That script will first copy to the clipboard whatever text was already selected.
  • Then it will see if the ChatGPT app is running, and if so, switch to it and focus the input box.
  • If it’s not running, it will launch it.
  • Then it’ll paste from the clipboard and append the word ‘explain’.

OK, open up the shortcuts.ahk file and paste in the following:

; Key shortcuts
; ^ = Control
; ! = Alt
; + = Shift
; # = Windows

; Ask ChatGPT about the selected text
^!+a:: ; ctrl + alt + shift + a
{
    Send("^c") ; Copy the selected text to the clipboard
    Sleep(100)

    ; If the ChatGPT window is open, activate it
    SetTitleMatchMode(3) ; Full match when using WinExist
    if (WinExist("ChatGPT")) {
        WinActivate
        Send("+{Esc}") ; Focus the input
        Sleep(100)
    } else {
        ; Else we'll have to open it
        ; Ctrl+Alt+Shift+Win opens Office 365! So wait for ctrl to be released
        KeyWait("Control")
        Send("#1") ; Where 1 is the taskbar position
        Sleep(2000) ; Wait for the app to load
    }

    Send("“^v” - explain")
}

A few things to note:

  • Comments start with semicolons. Gross.
  • You might need to fiddle with the Sleep values for your machine, they’re in milliseconds.
  • The folks over at Microsoft decided that ctrl + alt + shift + Win should open Microsoft Office 365 (one can only assume that mind-altering drugs were involved). AHK will trigger the Win + 1 key combo before the original shortcut keys (including ctrl, alt and shift) are released, meaning that Office 365 opens! So we wait for ctrl to be released before sending the Windows key.
  • You don’t need to use a shortcut to launch the ChatGPT window, you could also do something like Run(“C:\Users\USER_NAME\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Chrome Apps\ChatGPT”), pointing to the location where the ‘app’ was installed. I just figured if you’re reading this you’re kinda into shortcuts so would want it that way.
  • shift + esc is a shortcut that the ChatGPT folks have in their front end to focus the input box (stealing the shortcut from Chrome’s task manager). Other chatbots might automatically focus the text input box and not require this.
  • This script pastes the selected text in curly braces and appends the word ‘explain’. Having just a single word means that if I want to type something else here, I can press ctrl + shift + ← and type over the word ‘explain’, and if I want it to explain a particular word in a paragraph, I can just append that word.

To activate the above script (once saved), right click the AHK taskbar icon and select Reload Script. Or, if you’re using IntelliJ with the plugin, you can run the file from within the IDE (play button at the top left of the file) and that has the same effect. If it asks you if you want to replace the existing running script, click Yes.

Now to test that it’s working, select any text anywhere and hit ctrl + alt + shift + a and it should show up as a prompt in your chatbot of choice. If something goes wrong, fiddle with the sleep values.

Now, on to the second interaction…

Showing ChatGPT an area of the screen

In the same shortcuts.ahk file, paste the following below what you’ve already got:

; Show ChatGPT a screenshot
^!+s:: ; ctrl + alt + shift + s
{
    ; Ctrl+Alt+Shift+Win opens Office 365! So wait for ctrl to be released
    KeyWait("Control")

    Send("#+s") ; Win + shift + s to open Snip tool
    Sleep(100)

    ; Wait for user to capture screen area
    while !(GetKeyState("LButton", "P"))
        Sleep(10)
    while (GetKeyState("LButton", "P"))
        Sleep(10)
    Sleep(500)

    ; If the ChatGPT window is open, activate it
    SetTitleMatchMode(3) ; Full match when using WinExist
    if (WinExist("ChatGPT")) {
        WinActivate
        Send("+{Esc}") ; Focus the input
        Sleep(100)
    } else {
        ; Else we'll have to open it
        Send("#1") ; Where 1 is the taskbar position
        Sleep(2000) ; Wait for the app to load
    }
    Send("+{Esc}") ; Focus the input
    Send("^v")
    Send("explain")
}

A few notes:

  • This is mostly the same logic as the first script, but instead of sending ctrl + c to copy the selected text, it uses the built-in Windows shortcut Win + shift + s to capture an area of the screen. I’m pretty sure this always copies the image to the clipboard, but if not, open the Snip app and look for a setting to make it do this.
  • After calling this shortcut, we then need to wait for the mouse button to go down then up again while you draw a rectangle to select an area of the screen to capture.
  • Then we repeat the logic to focus ChatGPT and paste.

Save and reload the script, as before.

Now to test that it’s working, press ctrl + alt + shift + s, click and drag to select an area of the screen, and that image should show up in ChatGPT with the word ‘explain’.

I use this quite a lot. For example if I see some code in a YouTube video and want to run it, I’ll just press ctrl + alt + shift + s , draw a box around the code and tell ChatGPT to “type this out”. Same for any case of ‘images of text’ where I want the text, like URLs in a slideshow in a video (“make this a link”), or just pictures of interesting things where I want to know what the interesting thing is (“what’s on that person’s head?”). Oh and this is great for those error message windows with an error code that you can’t copy to the clipboard.

Now, there’s a bit of duplication in this code with the logic to ‘focus or open ChatGPT’. I don’t want to write a whole AutoHotKey tutorial, but I’ll leave you with a version of the full file that extracts the repeated code out into a function.

#Requires AutoHotkey v2.0-
; Key shortcuts
; ^ = Control
; ! = Alt
; + = Shift
; # = Windows

GoToChatGPT() {
    ; If the ChatGPT window is open, activate it
    SetTitleMatchMode(3) ; Full match when using WinExist
    if (WinExist("ChatGPT")) {
        WinActivate
        Send("+{Esc}") ; Focus the input
        Sleep(100)
    } else {
        ; Else we'll have to open it
        ; Ctrl+Alt+Shift+Win opens Office 365! So wait for ctrl to be released
        KeyWait("Control")
        Send("#1") ; Where 1 is the taskbar position
        Sleep(2000) ; Wait for the app to load
    }
}


; Ask ChatGPT about the selected text
^!+a:: ; ctrl + alt + shift + a
{
    Send("^c") ; Copy the selected text to the clipboard
    Sleep(100)

    GoToChatGPT()
    Send("“^v” - explain")
}


; Show ChatGPT a screenshot
^!+s:: ; ctrl + alt + shift + s
{
    KeyWait("Control")
    Send("#+s") ; Win + shift + s to open snip tool
    Sleep(100)

    ; Wait for user to capture screen area
    while !(GetKeyState("LButton", "P"))
        Sleep(10)
    while (GetKeyState("LButton", "P"))
        Sleep(10)

    Sleep(500)

    GoToChatGPT()
    Send("+{Esc}") ; Focus the input
    Send("^v")
    Send("explain")
}

; Bonus shortcut: search Google for the selected text
^!+x:: ; ctrl + alt + shift + x
{
    Send("^c")
    Sleep(100)

    Run("https://www.google.com/search?d&q=" . A_Clipboard)
}

Well, I hope that’s useful to at least one human out there, have a noice day.

Chatbots
AI
ChatGPT
Computers
Windows

--

--

David Gilbertson
David Gilbertson

Written by David Gilbertson

31K Followers
34 Following

I like machine learning stuff.

Responses (2)

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech