Turn on comment summaries in your ChatGPT workspace
Before you start trying to auto summarize comments from Google Docs, go check if your ChatGPT workspace even lets you do that. As of right now, this flows through the Advanced Data Analysis (formerly Code Interpreter) feature, but it still only works if you manually share the Doc’s export OR pull it via the API. There’s no native “connect my Google Docs” yet — and believe me, I keep checking like every two weeks 😅
If you’re using ChatGPT Plus, click the little gear next to your name, go to Settings > Beta Features, and make sure Advanced Data Analysis is switched on. If that’s not there, it probably means you’re inside an organization account and the admin disabled it. Ask someone nicely… or use your personal account instead ¯\_(ツ)_/¯
Export the comments from Google Docs manually
This is still the most repeatable way I’ve found. I’ve tried asking ChatGPT to fetch comments from a shared Google Doc link — even giving it editor access — and it just fails silently. No errors. No helpful suggestions. It just… pretends the link is blank. So now I just export comments manually.
To do that:
1. Open the Doc with comments turned on.
2. Go to File > Download > Microsoft Word (.docx)
3. Save the file and remember the location 😅
Now here’s the thing: Only the .docx version keeps the comment threads in a format that ChatGPT can read. PDFs flatten everything. Copy-pasting loses the author, timestamps, and structure. And good luck parsing anything from Google Doc’s JSON exports unless you love pain.
You could also select File > Email > Email collaborators and choose “Include comments.” That sends them as a flat list in the email body, but I found that doesn’t work great if you have more than maybe 20 or 30 comments. It just becomes spaghetti.
Drag and drop the Word file into ChatGPT
So here’s where the magic does happen — if the model behaves. Open a new chat with ChatGPT and switch the mode to Advanced Data Analysis. Then just drag the Word file into the chat box.
I usually lead with a prompt like this:
“This file contains Google Doc comments. Please summarize the feedback by:
• Grouping similar comments by topic
• Highlighting any conflicts or request-for-clarification threads
• Noting how many total comments there are, and how many are resolved versus unresolved”
If the document’s small, it does all this in a minute or less. But sometimes it’ll just say “This appears to be a Word document. How can I help with it?” and not even notice the comments. In that case, rephrase your request more directly — something like “List all comment threads in this file by commenter, location, and content.” That usually triggers it to parse the annotations properly.
It’s weird how sometimes it’ll get stuck on a sentence and try to summarize that instead of the actual comments. You can always type “Try again, but only summarize the comments.” That usually resets it.
Fallback method using Google Apps Script to extract comments
If Word export isn’t giving you what you need — maybe because there’s formatting issues, or thread positions aren’t helpful — here’s something I set up as an escape hatch. It’s ugly but it works 🤷♀️
1. Open your Google Doc
2. Click Extensions > Apps Script
3. Paste in this snippet:
“`javascript
function exportComments() {
var docId = DocumentApp.getActiveDocument().getId();
var comments = Drive.Comments.list(docId);
var output = ”;
for (var i = 0; i < comments.items.length; i++) { var c = comments.items[i]; output += 'Author: ' + c.author.displayName + '\n'; output += 'Date: ' + c.createdDate + '\n'; output += 'Content: ' + c.content + '\n'; output += 'Resolved: ' + c.status + '\n'; output += '---\n'; } Logger.log(output); } ``` 4. Run the script (you’ll need to enable Drive API in the Apps Script dashboard → Services). 5. The result shows up under View > Logs.
Copy that log output and paste it into ChatGPT with the prompt:
“Here are all the Google Doc comments. Group by topic and summarize suggested changes.”
It doesn’t have positions this way, but at least you get every comment cleanly.
How I built an auto summarized daily report from a shared folder
I was loosely updating a running draft every couple of days, collecting feedback and tag suggestions inside the doc itself. Eventually I set up a Google Drive folder where every file would get dumped — and I wanted to summarize all comments every night and email them to myself.
Now here’s the fun (aka pain): Google Apps Script can’t pull Comments from docs created by another user unless you own them, even if they’re in a shared folder. And Zapier or Make don’t expose the comment text either.
So I added a Google Workspace account that would act as the official owner of the folder, then created a scheduled Apps Script like:
“`javascript
function summarizeFolderComments() {
var folder = DriveApp.getFolderById(‘FOLDER_ID’);
var files = folder.getFiles();
while (files.hasNext()) {
var file = files.next();
var comments = Drive.Comments.list(file.getId());
// Build a summary
}
// Email yourself the summary
}
“`
And then pushed every output to a Google Sheet, which I formatted as a summary table ChatGPT could read:
| File Name | Comment Count | Topics | Open Threads |
|———–|—————|——–|————–|
| roadmap.ai | 12 | onboarding, UI friction | 4 |
| draft_v2 | 7 | copy changes, tone | 2 |
Then I drag the CSV into ChatGPT and ask it to make an executive summary. It’s blunt, but it gets the job done. I’m still tweaking how it recognizes repeating feedback strings though, because sometimes threads that say totally different things get grouped together 🤔
Use Claude if ChatGPT keeps missing the comments
Here’s something real: Sometimes Claude from Anthropic just reads Word comments better than ChatGPT. I don’t fully understand why — I think it might parse the file’s comment metadata a different way — but I’ve had six documents in a row that ChatGPT didn’t summarize properly, but Claude nailed.
Drag the same Word file into Claude’s chat and say:
“This file includes Google Doc feedback. Please group the comments by theme and summarize the key changes requested.”
It feels more stable with large comment threads. It also doesn’t hallucinate missing comments the same way ChatGPT sometimes does — I once had a 23-comment file get summarized as having 11. I cross-checked every one.
Downside? Claude’s file size limit is tighter with the free tier. But if you’re just working with under-50-page Docs, it’s worth trying.
Summarizing comment threads during live editing sessions
If your document is still being edited and people are leaving fresh comments, exporting and importing every time is a pain. What I’ve done in short sessions (like during client workshops) is:
• Keep the doc open with “comment only” view
• Use the requestComments() method in the Google API via Apps Script OR Postman
• Pipe each comment into a local text file, then paste that into GPT
It’s not real-time, but it’s quick enough for feedback rounds that happen live. And the real trick is turning each comment into a bullet under the section it belongs to. ChatGPT can usually infer the section if the comment says something like:
“We should soften this intro to feel more collaborative”
But it can’t if all you give it are isolated comment texts. You’ll need to manually jog its memory by showing which part of the paragraph the comment refers to.
Honestly, this is the messy part I still haven’t fully solved 😵💫
Verbally review ChatGPT summaries before acting
This is slightly embarrassing, but I once acted on a summarized comment about “tone being more neutral” — except it had reversed who even said it. Took me 3 hours to realize ChatGPT accidentally attributed the comment to the wrong person in a thread.
So now I read each summary aloud while looking at the original doc, and I cross-reference the commenter name and context. If anything seems off — like a comment about spelling tagged as “editorial feedback” — I dig into the raw version.
ChatGPT is helpful, but it sometimes collapses nuance. Especially with responses like:
• “Several commenters requested clarity in section two” — when it’s actually two people arguing about what clarity means
• “Minor formatting issues noted” — which turns out to be a debate about bullet vs numbered lists across the whole thing
Summarizing comments saves time, yes. But only if you sanity-check what it says.