What a ChatGPT share link contains

A shared conversation link is a view of selected ChatGPT content. Current conversation links start with https://chatgpt.com/share/.

OpenAI says that for personal accounts, anyone with the link can view the shared content. The link shows the content available when it was created or last updated; later messages are not added automatically. Links created in managed workspaces may be limited to eligible members of that workspace. See the OpenAI Help Center guide to shared links.

Think of a personal share link as a snapshot that can be forwarded, not as a private document or a live window into the owner's account.

Two checks before you convert a link

Check the contents. A personal link can be opened by anyone who gets it, and there are no recipient-by-recipient access controls. Read the shared preview first and remove names, account details, API keys, or other private information before creating a link. If someone sent you the link, make sure they intended the conversation to be shared with you.

Remember that links and files can be copied. The owner can update or delete a shared link. Deleting it stops future access through that URL, but it does not remove a copy another person already imported into their ChatGPT history. A PDF or Word export is also a separate file that can be forwarded, so store it with the same care as the conversation.

Choose a way to save it

Print the page to PDF

For a short, mostly text conversation, open the share link and use your browser's Print → Save as PDF. It is quick and needs no extra tool. Long chats, wide tables, code, and equations can break across pages or lose formatting, so review the result before sending it on.

Copy it into a document

For a few short messages, copying the visible text into Word or Google Docs gives you control over the final layout. Tables and formulas may need manual cleanup.

Convert the share link

If the conversation has tables, code, or math, a converter can save the cleanup work. With TODOC's ChatGPT share link converter, paste a public share URL, load the conversation into the editor, remove turns you do not need, then export to PDF, Word, or Excel. The converter currently accepts ChatGPT share links only.

TODOC processes the conversion on its server. Its privacy policy says source content is not retained separately after conversion and is not used to train AI models; exported files are kept according to the export-history period for the selected plan. Check that policy before using the service with work or client material.

A simple workflow

1. Open the shared conversation and review what it contains.
2. Choose Print for a short, simple chat, or use a converter when tables, code, or equations need to remain readable.
3. Review the exported file, then keep it somewhere with the right access controls.

If the link does not open, ask the owner to check whether the chat or link was deleted, or to create or update a link. The OpenAI shared-link guide explains how owners manage those links.

Disclosure: I'm the developer of TODOC, the converter mentioned above. Printing or copying a conversation works without it.

This guide shows how to convert Markdown to Word properly, mostly with pandoc, and covers the problems I ran into after doing it a lot.

The one-liner

If you have pandoc installed, a basic Markdown to Word conversion is one command:

pandoc notes.md -o notes.docx

Headings become real Word heading styles (so the navigation pane and table of contents work), lists become real lists, and pipe tables become Word tables. For a plain document, you can stop here.

Make it look like your documents: the reference doc

The default output uses pandoc's built-in styles. To match your company or school template, generate the reference file, edit its styles in Word, and pass it back:

pandoc -o reference.docx --print-default-data-file reference.docx
# open reference.docx in Word, change the styles (Heading 1, Normal, Table...), save
pandoc notes.md --reference-doc=reference.docx -o notes.docx

Only the styles in reference.docx are used, not its content. If a change doesn't show up, you probably edited the text formatting directly instead of modifying the style.

Math: get real Word equations, not images

This is where most Markdown to Word converters fail. Pandoc's docx writer turns TeX math into native Word equations (OMML), so the equation stays editable in Word. You just have to make sure pandoc recognizes the math first.

Pandoc's Markdown understands $...$ and $$...$$ by default. But a lot of Markdown in the wild, including what you copy out of ChatGPT, uses \(...\) and \[...\]. Enable that explicitly:

pandoc -f markdown+tex_math_single_backslash notes.md -o notes.docx

Two gotchas worth knowing:

1. Dollar signs that are prices. Pandoc only treats $...$ as math when the opening $ has a non-space character right after it, and the closing $ has a non-space character right before it and is not followed by a digit. That rule keeps It costs $5 and $10 as text, but documents that mix prices and formulas on the same line can still trip it up. The safest habit is to escape prices as \$5.
2. Display math inside lists. $$ blocks indented inside a list item sometimes get parsed as part of the paragraph. Put a blank line before and after the block.

Tables

Pipe tables convert fine:

| Plan    | Price | Change |
|---------|------:|-------:|
| Basic   |  $9   |   +4%  |
| Pro     |  $29  |  -2%   |

Things to watch:

• Column widths. Pandoc lets Word auto-size the columns. Wide tables with long text can overflow the page; set widths in your reference doc's table style, or fix them by hand in Word.
• Line breaks in cells. Pipe tables can't hold multi-paragraph cells. Use <br> for a line break, or switch to pandoc's grid tables if you need real paragraphs.
• Numbers. In Word a number in a table is just text. If the reader is going to sum the column, they want Excel, not Word.

Code blocks

Fenced code blocks with a language tag keep a monospace font and get syntax highlighting in the .docx:

```python
def growth(start, end, years):
    return (end / start) ** (1 / years) - 1
```

Always add the language tag. Without it you get a plain monospace block, which is fine but harder to read.

Mermaid diagrams

Pandoc doesn't render Mermaid. Inside the .docx you'd just see the source code. Render diagrams to images first with mermaid-cli, then reference the image:

npx -p @mermaid-js/mermaid-cli mmdc -i flow.mmd -o flow.png

![Signup flow](flow.png)

Tall diagrams (vertical flowcharts) can run off the bottom of the page. Add a height attribute so pandoc scales them by height: ![Signup flow](flow.png){height=8in}.

A full command

Putting it together:

pandoc notes.md \
  -f markdown+tex_math_single_backslash \
  --reference-doc=reference.docx \
  -o notes.docx

The no-install option

If you don't want to set up pandoc, or you need to hand this to non-developer colleagues, I built a browser tool for exactly this: a Markdown to Word converter where you paste Markdown, check the preview, and download a .docx. It runs the same kind of pipeline on the server, with the math delimiter handling, table sizing and Mermaid rendering above already done for you. Your first export doesn't need an account.

Checklist

• -f markdown+tex_math_single_backslash if your source uses \( \)
• Escape $ when it means money
• Blank lines around $$ blocks
• Language tag on every code block
• Mermaid rendered to PNG before conversion
• --reference-doc for house styles

What breaks for you when you convert Markdown to Word? I'm collecting edge cases, so drop them in the comments.

Disclosure: I'm the developer of TODOC, the converter mentioned in the "no-install option" section. Everything else in this post works with plain pandoc.

Here is a practical workflow for preparing a product image with ZeroBG, an AI background-removal tool.

1. Start with a usable source image

Keep an untouched copy of the original. A sharp image with clear separation between the product and its background usually gives any automatic remover more useful information. If the item blends into a similar-colored surface, take a second photo with more contrast or use a different angle when possible.

Avoid cropping tightly before the cutout. Leaving a little room around the item makes it easier to inspect the result and choose a final composition later.

2. Generate a cutout preview

ZeroBG accepts JPG, PNG, and WebP images. Drop in a photo and review the background-removal preview before deciding whether to keep the result. The preview is free and does not require an account; it is limited to 512 pixels on the longest side and includes a watermark.

For a reusable asset, a transparent PNG is often the most flexible output. It can sit on a white marketplace background today and a campaign color tomorrow. If the image is only for a listing, a solid white background may be quicker.

3. Inspect the details that are easy to miss

Zoom in around thin edges, handles, reflective surfaces, and small gaps. Hair and fur need soft edges; glass and translucent objects need enough detail to avoid looking like opaque cutouts. A clean preview at thumbnail size can still reveal rough edges when viewed at full size.

ZeroBG includes a touch-up brush for erasing leftover background or restoring parts of the subject. Automatic removal is not perfect on every photo: a white product on a white table, a very busy background, or motion blur may need a different source image or manual correction.

4. Choose a background for the destination

Use transparent PNG when the same cutout needs to work across several layouts. Use a solid background when a marketplace or design template requires one. Keep the shadow and spacing consistent across a set of products so the images look like one catalog rather than a collection of unrelated photos.

Before export, check the crop, dimensions, and color of the background against the destination where the image will appear. A subtle mismatch between “white” and the page’s actual white can be noticeable in a product grid.

5. Keep the original and the final asset together

Save the original, the approved cutout, and any final-background version with clear filenames. That small habit makes future edits easier and prevents a compressed listing image from becoming the only copy you have.

ZeroBG offers free watermarked previews; full-size, watermark-free downloads use credits. Check the ZeroBG site for current export options and pricing.

The useful workflow is simple: start with a clear photo, generate a cutout, inspect the edges, and export for the actual destination. Background removal can save tracing time, while a final human check keeps the product looking like itself.