Working as of 16/11/2021
If this script helped or interested you, please consider staring the repo above. That number looks cool when it's big
- Download a userscript manager like TamperMonkey for Chrome or Greasemonkey for Firefox.
- Use this link to install the script.
- Click the extension while in Khan Academy and ensure both the extension and the script are on.
- Open Developer Tools and go to the Console tab. The script will console log answers as the browser gets them.
- Khan Academy always requests the current and next question, so expect the second to last console log message to be the correct answer
- When there are multiple answers, fill in the boxes left-to-right and then down. You can also press tab to get to the next field. Example below
- Change your console log level to only info for a much better experience
- This works only for expression, free response, multiple choice, and dropdown questions.
- The script will do its best to find the answer in the question data, but some edge case questions do not follow the same structure, therefore I can't account for them. (read below to understand the exploit)
- If the script breaks for a certain type of question, open an issue and I will take a look.
- I am lazy, therefore this is buggy. Don't be surprised when you run into a question you'll actually have to do
It's pretty simple. On every 'quiz' you open in the app, your browser makes a request to /getAssessmentItem
. The server responds with everything your client needs to draw and grade your question. Within this graphql response is a json blob containing a list of questions, most of them with a correct: boolean
attribute. Sometimes, they try to throw you off by placing correct
as the value to a key labeled either status
or considered
.
I wrote this for Chrome, although all should work on gecko. It essentially hooks into the browser's fetch
, which is what Khan Academy uses now instead of XMLHttpRequest
(this is why the some old exploits no longer work, along with the endpoint change), and when /getAssessmentItem
is requested, it logs the "important" part of the response.
This is fully open source! If you want to take a crack at automating this script, or if you'd like to add support for more types of questions, like matching for example, feel free to submit a pull request. I'll be sure to credit you accordingly!
This has the GNU GPL 3.0 licence. I expect most users will be people like me who have AP Stats assignments and no will to do them. I don't care too much about what you do with it, but I like credit :)