CopyCharm
CopyCharm
← Back to blog

How to Fix Garbled Characters in Exported AI Conversations

Summary

  • Garbled characters after exporting AI chats usually come from a text-encoding mismatch (UTF-8 vs Windows-1252/ANSI) or a font that cannot display certain symbols.
  • First identify the symptom (mojibake like “’” or replacement boxes “�”) and the file type (TXT, HTML, JSON, PDF, DOCX) to choose the right fix.
  • The most reliable repair is to re-open the export with the correct encoding (UTF-8) and re-save, or re-export using a UTF-8-safe path.
  • App-specific steps differ: Windows Notepad, macOS TextEdit, Excel, Google Docs, and Word each have different import/encoding behaviors.
  • Always do a result check using a short “known-bad” test string and a quick search for common corruption patterns before archiving or sharing.

Exported AI conversations can look fine inside the chat app, then turn into unreadable symbols when you open the exported file elsewhere. This is frustrating because it feels random: quotes become “’”, em dashes become “—”, emoji become boxes, and non‑Latin characters turn into “????”.

The good news: in most cases, the text is still there. It is being interpreted with the wrong character encoding (or displayed with a font that cannot render it). This guide walks you through a reproducible example, then gives application-specific fixes for the most common places people open exported AI chats (Notepad/Word/Excel/Google Docs), with a clear result check at the end.

What “garbled characters” usually mean (so you pick the right fix)

Different symptoms point to different root causes:

  • Mojibake (encoding mismatch): You see sequences like ’, “, —, é. This often happens when UTF-8 text is opened as Windows-1252 (ANSI) or ISO-8859-1.
  • Replacement character: You see � (a black diamond with a question mark). This indicates the app could not decode some bytes with the chosen encoding, or the text was already damaged during a conversion.
  • Question marks: You see ???? where non-ASCII characters should be. This can happen when saving to an encoding that cannot represent those characters (for example, saving as “ANSI”).
  • Boxes/tofu: You see empty squares for emoji or some scripts. The text may be correctly encoded, but the font used to display it lacks glyphs.

A reproducible example (so you can confirm the fix worked)

Use this short test string to reproduce and verify repairs. It includes curly quotes, an em dash, accented characters, CJK, and emoji:

  • Test string: “Curly quotes” — apostrophe ’, café, naïve, 中文, 日本語, 😀

How to use it: paste the test string into a new chat message (or into a note you plan to export alongside your conversation), export the conversation, then open the exported file in the app where you see corruption. After applying a fix below, re-open and confirm the test string displays exactly as written.

Step 1: Identify the export format and where corruption is introduced

Before changing anything, answer these two questions:

  • What file type did you export? Common ones are .txt, .html, .json, .pdf, or a document format like .docx.
  • Where does it look wrong? Is it wrong in a browser preview, in Notepad, in Word, in Excel, or only after copy/paste into another tool?

Quick isolation check: open the exported file in two different viewers. If it looks correct in one but broken in another, the export is probably fine and the issue is how the second app is importing/decoding it.

Step 2: Fix the most common cause: open and re-save as UTF-8

If you see mojibake like “’” or “é”, start here. The goal is to open the file using the correct encoding (usually UTF-8) and then re-save it in UTF-8 so other apps interpret it correctly.

Windows: Notepad (Windows 10/11)

Best for: TXT, CSV, JSON, HTML that opens as plain text.

  1. Right-click the exported file and choose Open with > Notepad.
  2. If the text is garbled, try: File > Save As.
  3. In the Save As dialog, look for an Encoding dropdown.
  4. Choose UTF-8 (not “ANSI”).
  5. Save a new copy (do not overwrite yet), then re-open the new file.

Result check: search within the file for ’ or Ã. If those sequences disappear and curly quotes/accents look correct, you fixed an encoding mismatch.

macOS: TextEdit

Best for: TXT exports opened on Mac.

  1. Open TextEdit.
  2. Use File > Open and select the exported file.
  3. If it opens in rich text mode and looks odd, switch to plain text: Format > Make Plain Text (this can change formatting; use a copy of the file).
  4. Use File > Save (or Duplicate first) and keep it as plain text.

Note: TextEdit’s encoding controls vary by macOS version and file type. If you cannot choose encoding, use a different editor that allows explicit UTF-8 selection.

Browser vs desktop note (important for HTML exports)

If your export is .html and it looks garbled in a browser, the file may be missing a UTF-8 declaration. A browser relies on metadata to decide how to decode the page.

What to look for: open the HTML in a text editor and check the first ~20 lines for a charset declaration such as:

  • <meta charset="utf-8">

If it is missing, adding it can fix display in browsers. Save a copy of the HTML, add the meta tag inside the <head> section, then reload the file in the browser.

Step 3: App-specific fixes (where exports commonly break)

Microsoft Word (desktop)

Common failure mode: Word opens a text file but guesses the wrong encoding, or you paste content and Word substitutes characters.

  • Try “Open and Repair” style import: Instead of double-clicking the file, open Word first, then use File > Open and select the file. If Word offers any text encoding/import options, choose UTF-8.
  • Prefer inserting text over pasting: If copy/paste introduces corruption, try inserting the file content via Word’s insert options (Word may treat it differently than clipboard paste).
  • Font check: If emoji or some scripts show as boxes, change the font for that section to a font with broader Unicode coverage.

Result check: confirm curly quotes stay curly, and that accented characters (e.g., “café”) remain intact after saving as DOCX and reopening.

Microsoft Excel (desktop) and CSV exports

Common failure mode: Excel imports CSV/TXT using a legacy encoding guess, which can corrupt non-ASCII characters.

Safer approach: import the file through Excel’s data import flow rather than opening it directly. Look for an import path that lets you choose:

  • File origin / encoding (choose UTF-8)
  • Delimiter (comma, tab)

Result check: verify that non-English characters and emoji remain correct in cells, then save as an Excel workbook format to preserve them.

Google Docs (browser)

Common failure mode: the upload is fine, but copy/paste from a source app introduces hidden encoding/formatting issues.

  • Upload the file instead of pasting: If you have a TXT/HTML export, upload it to Drive and open with Docs.
  • Paste without formatting: If pasting is required, use Docs’ “paste without formatting” shortcut to reduce weird transformations, then re-apply formatting manually.
  • Check for smart quotes conversions: If you need plain ASCII quotes for code, turn off smart punctuation in your workflow by pasting into a plain-text editor first.

Result check: search the document for � and for sequences like â€. If present, the corruption happened before or during import.

PDF exports

Common failure mode: the PDF looks fine, but copying text out of it produces garbled characters due to how the PDF encodes text internally.

  • Prefer exporting to HTML or TXT for reuse: If your goal is to reuse the conversation text, a text-based export is easier to keep clean.
  • If you must use PDF: try copying from a different PDF viewer, or use the export’s original source format if available.
  • Font embedding: If the PDF displays boxes, it may be missing fonts for certain glyphs. Re-exporting with font embedding (if the exporter offers it) can help.

Result check: copy a paragraph containing curly quotes and emoji into a plain-text editor. If it breaks only after copy, the PDF is the bottleneck.

Step 4: Fix garbling caused by double-encoding (common “’” pattern)

If you see ’ where you expected ’, that is a classic sign that UTF-8 bytes were interpreted as Windows-1252. The fix is not to “replace characters” manually across the whole file first; it is to re-open the file with the correct encoding and re-save.

When manual replace is acceptable: If you only have a few instances and cannot re-export or re-open with the correct encoding, you can do targeted replacements (for example, replacing “’” with “’”). Do this only after making a backup, and run a result check to ensure you did not introduce new errors.

Step 5: Prevent the problem on your next export

  • Prefer UTF-8-safe formats: If you can choose, export to a format that preserves Unicode cleanly (TXT/HTML/JSON with UTF-8) rather than a legacy encoding.
  • Avoid “ANSI” saves: On Windows, “ANSI” can drop characters it cannot represent, turning them into “?” permanently.
  • Keep a “known-bad” test string: Include the test string from this article in a small note so you can quickly validate any export pipeline.
  • Be careful with intermediate apps: Corruption often happens when you open a file in one app and re-save it in a different encoding.

Final result check (do this before you archive or share)

  1. Re-open the fixed file in the app where it was previously garbled.
  2. Find the test string (or a line with curly quotes, em dashes, accents, and emoji).
  3. Confirm:
    • Curly quotes display as “ ” and ’ (not “ †or ’).
    • Accents display correctly (café, naïve).
    • Non-Latin text displays correctly (中文, 日本語).
    • No replacement characters (�) appear.
  4. Open the same file in a second viewer (for example, a browser and a text editor) to confirm it is broadly compatible.

Frequently Asked Questions

FAQ 1: Why do exported AI conversations show characters like “’” and “é”?
Answer: Those patterns usually indicate an encoding mismatch: the text was saved as UTF-8, but the app opening it interpreted it as a different encoding (commonly Windows-1252/ISO-8859-1). Re-opening the file with UTF-8 and re-saving in UTF-8 usually resolves it.
Takeaway: “’” is a decoding problem more than a typing problem.

Back to FAQ Table of Contents

FAQ 2: What is the fastest way to fix garbled characters in a .txt export on Windows?
Answer: Open the file in Notepad, use File > Save As, and choose UTF-8 in the Encoding dropdown, saving to a new file name. Then re-open the new file and confirm the garbled sequences are gone.
Takeaway: Re-save as UTF-8 (not ANSI) to preserve Unicode characters.

Back to FAQ Table of Contents

FAQ 3: How do I fix garbled characters in an HTML export when it looks wrong in a browser?
Answer: Open the HTML file in a text editor and check whether it declares UTF-8 (for example, a meta charset tag). If it is missing, add <meta charset="utf-8"> inside the head section, save a copy, and reload it in the browser.
Takeaway: Browsers rely on charset metadata to decode HTML correctly.

Back to FAQ Table of Contents

FAQ 4: Why does the export look fine, but copying from a PDF makes the text garbled?
Answer: PDFs can store text in ways that display correctly on-screen but copy out poorly, depending on the viewer and how the PDF was generated. If you need reusable text, exporting to a text-based format (TXT/HTML/JSON) is usually more reliable than copying from PDF.
Takeaway: A PDF can be visually correct while still being copy-unfriendly.

Back to FAQ Table of Contents

FAQ 5: How can I prevent Excel from mangling characters when importing a chat export?
Answer: Avoid opening CSV/TXT exports by double-clicking. Instead, import through Excel’s data import flow and select UTF-8 as the file origin/encoding if the option is available, then confirm special characters render correctly before saving as a workbook.
Takeaway: Use an import flow that lets you choose UTF-8.

Back to FAQ Table of Contents

FAQ 6: What does the “�” replacement character mean, and can it be reversed?
Answer: “�” appears when an app cannot decode some bytes using the selected encoding, or when characters were lost during a prior save/conversion. If the underlying bytes are still intact, opening with the correct encoding can restore the text. If the file was saved in a way that replaced characters with “?” or “�”, the original characters may not be recoverable from that file alone.
Takeaway: Sometimes it is reversible, but not if the file was already overwritten with loss.

Back to FAQ Table of Contents

FAQ 7: Is this a font problem or an encoding problem?
Answer: If you see sequences like “’” or “Ô, it is usually encoding. If you see empty squares/boxes while the rest of the text looks normal, it is often a font that cannot display those characters (commonly emoji or certain scripts). Try changing the font to test; if the characters appear, the encoding was likely fine.
Takeaway: Mojibake points to encoding; boxes point to fonts.

Back to FAQ Table of Contents

FAQ 8: What should I do if the text is already corrupted in the exported file itself?
Answer: First, confirm by opening the export in at least two different viewers. If it is corrupted everywhere, re-export the conversation using a different format (TXT/HTML/JSON) if available, and avoid any “ANSI” or legacy encoding save steps. If you cannot re-export, restore from an earlier uncorrupted copy if you have one, then apply the UTF-8 re-save workflow.
Takeaway: If the export is corrupted at the source, re-exporting in a Unicode-safe format is the cleanest fix.

Back to FAQ Table of Contents

CopyCharm for AI Work
Turn copied work snippets into clean AI context.
CopyCharm helps you turn copied work snippets into clean, source-labeled context packs for ChatGPT, Claude, Gemini, Cursor, and other AI tools. Copy, search, select, and export the context you actually want to use.
Download CopyCharm

Related Guides