How to create an AI writing coach with OpenAI's GPT and Zapier

Like any skill, writing takes practice. Some people have a natural knack for stringing words together, but many writers look for ways to improve their craft over time. 

But feedback isn't always readily available in the furious world of online writing. Likes, comments, and shares might be vague indicators, but don't really tell you how you can improve. Many writers work without the collaboration of an editor, and it can be hard to critique your own work.

But what if you had your own personal writing coach—one with access to billions of writing samples who could offer you straightforward suggestions for improvement? AI can act as that coach. All it takes is a combination of writing samples from you and the right AI prompt.

How can AI give you writing feedback?

If you hand a writing sample to a friend, a colleague, and an editor and say, "Give me feedback," you're going to get three very different responses. Each will take a different approach.

AI generates a response based on your written instructions, or prompts. The more specific you are with your prompt, the better your results will be. 

You can ask AI to point out ways you can improve, with instructions like: 

• Provide constructive criticism

• Ask how the writing sample could be improved

• Describe the sample using three adjectives

• Give examples of improvements to encourage the writer

You could even get more creative and ask the AI to criticize the article as a Ph.D. of literature or an editor for a famous publication. It depends on how thick your skin is…

But the important part is to track the feedback. If you get into a daily habit of writing a short sample, you can review the critiques from the AI over time and look for patterns. Does the AI consistently say you're too wordy? Not imaginative enough? You can use the feedback to improve your writing samples over time. 

Build a writing coach with Zapier and OpenAI's GPT

You can systemize your flow of inputting writing samples and tracking the AI feedback with Zapier. 

You'll need to create a simple form for input (such as Google Forms), connect OpenAI's GPT-3, and output the results to a format of your choice, such as Notion, an email, a GSheet, or other tools. 

If you'd like to start with a template, click on the Zap template below, and you'll be taken to the Zap editor. You'll need to create a Zapier account if you don't already have one. Then, follow the directions below to set up your Zap.

Step 1: Connect your OpenAI account to Zapier

To use OpenAI's language model (GPT-3), you'll need to first create an OpenAI account. 

1. Within OpenAI, generate an API key. Copy the key.

2. In Zapier, click on My Apps, then + Add Connection.

3. Search for OpenAI.

4. Paste the API Key from your OpenAI account.

Step 2: Build a form to input your writing sample

Your form could be very simple and only include a text box if you plan to send the same type of writing sample and want the same type of feedback each time. 

But you could also build a more robust AI writing coach by including fields that specify the type of writing sample and the type of feedback you want. This is particularly useful if you write in several different formats, such as blog posts and short stories. 

The prompts and feedback types can be built as a dropdown selection or radio buttons. 

Then you'll create a text box on your form to add your writing sample.

Step 3: Set up your Zapier trigger to pull in your form response

A Zap always begins with a trigger, the event that starts your Zap.

To start creating this Zap: 

1. Select your form app (Google Forms in this case)

2. Select your trigger event—we'll go with New Form Response.

3. Select the form you've created.

Step 3: Set up your OpenAI action to send your prompt

Now you need to set up the prompt you want the Zap to send to GPT-3. Your prompt will be a combination of static text (the same every time the Zap runs) and your form's responses.

If your form only includes your writing sample and no additional selections, then your prompt might be something like: You are a writing coach. Given the following text, what are your suggestions for improvement? [writing sample].

If your form includes a description of the type of text and/or the type of critique you want, then your prompt would be more like: You are a writing coach. The writer has been instructed to [type of writing prompt]. You should [type of writing feedback].Here is the writing sample: [sample].

The completed prompt with the text from your form would read like this:

You are a writing coach. The writer has been instructed to describe a setting in one paragraph. You should describe how the writing sample could be improved. Here is the writing sample [sample].

To set up this action step in your Zap.

1. Select OpenAI for your action app.

2. Select Send Prompt for your action event.

3. In the Prompt field, add your prompt text and map the fields from your form by selecting them from the data dropdown.

In the action step, you can also indicate the Maximum Length of OpenAI tokens to be used in the response. 

OpenAI uses tokens to determine the length of the output (and bills you according to the number of tokens you use if you're on a paid plan). Type in a lower number if you want a more concise response from your AI writing coach or want to use fewer OpenAI tokens. 

Step 4: Set up your final action to output the results

The OpenAI response to your prompt will be generated in Zapier, so your final step is to output it somewhere where you can review the results. 

Think about the best way you learn and how you'd like to store and review the information in the future.

• Would you like all of the original prompts and responses gathered in a tool like Notion so you can add your own notes?

• Would you like the response emailed to you?

• Do you want to keep the responses with the original prompt in a GSheet?

Heck, you could even send yourself a Slack message. The possibilities are endless, and you can create more than one action step with your OpenAI results. 

For this example, we'll create a Notion page with the original prompt and the OpenAI response.

To set up this action step:

1. Select Notion for the action app.

2. Select Create Page (or Create Database Item) for the action event.

3. Add the data from your previous steps to add your writing sample from your form step and your feedback from the OpenAI action step to the Content field.

Note: Choices Text has now changed to Response in the dropdown menu.

You could even use the form date as part of the page Title (and use a Zapier Formatting step if the form's date output is ugly).

After this step has been completed, you can review your newly created page in Notion.

You can make notes or add tags to your output—and that should be the driver behind the output format you choose (such as organizing your feedback in Notion or adding categories in a GSheet). 

Or maybe you simply send yourself an email with the OpenAI results, read and absorb the feedback, and then send it straight to archive. 

Review your results and tweak your prompts

As you review the feedback (with a grain of salt, of course, since it's a robot and not a human editor!), you can gradually change your writing samples over time. Or maybe you tweak your initial sample later that same day or a week later and try again, reviewing the new feedback from your AI writing coach. 

Once you feel like you've mastered a particular skill, tweak the prompts and types of feedback in your form if you'd like to work on something else.

Or, maybe you've "graduated" from an AI-coaching school—hopefully satisfied with the improvements you've made to your writing.