ChatGPT vs Claude vs Gemini: Which AI Has the Best Formatting?

All three major AI tools, ChatGPT, Claude, and Gemini, output markdown. But the quality and structure of that markdown varies. If you're choosing an AI tool partly based on how well its output pastes into documents, here's what you need to know.

The test

We asked each AI tool the same question: "Create a project status report for a software project with 4 workstreams, including a summary table, key risks, and next steps."

We compared the responses across:

• Heading structure
• Table formatting
• List organization
• Code block usage
• Overall readability when pasted

ChatGPT (GPT-4o)

Heading structure: Consistent use of H2 for major sections and H3 for subsections. Rarely uses H1 in responses.

Tables: Well-formed with aligned columns. Uses bold in header rows. Tends to include helpful columns like "Status" with values like "On Track" or "At Risk."

Lists: Uses both bullet points and numbered lists appropriately. Good nesting for sub-items.

Formatting habits:

• Tends to be verbose with more sections and detail
• Frequently uses bold for key terms within paragraphs
• Includes horizontal rules (---) between major sections
• Sometimes over-formats with excessive bold

Overall: Produces the most structured output. Sometimes too structured; the formatting can feel heavy for simple questions.

Claude (Sonnet/Opus)

Heading structure: Clean hierarchy with H2 and H3. More conservative than ChatGPT about adding sections.

Tables: Well-formed and concise. Tends to create simpler tables with fewer columns. Cell content is brief.

Lists: Excellent nesting. Uses indentation consistently. Prefers bullet lists over numbered lists unless order matters.

Formatting habits:

• More concise overall, less formatting "noise"
• Uses bold sparingly and purposefully
• Better at matching formatting to the complexity of the question
• Rarely produces unnecessary horizontal rules
• Uses blockquotes occasionally for important callouts

Overall: Produces cleaner, more readable markdown. Less likely to over-format. The output pastes well because the structure is proportional to the content.

Gemini

Heading structure: Similar to ChatGPT with H2 and H3, but occasionally inconsistent in heading levels within a response.

Tables: Sometimes includes extra whitespace in cells. Alignment can be inconsistent. Generally functional but less polished than ChatGPT or Claude tables.

Lists: Good basic list support. Nested lists sometimes have formatting inconsistencies.

Formatting habits:

• Can switch between formatting styles within a response
• Sometimes mixes heading levels unexpectedly
• Uses bold frequently, similar to ChatGPT
• Occasionally outputs formatting that doesn't follow strict markdown spec

Overall: Produces usable markdown but with occasional inconsistencies that can affect paste quality.

Side-by-side comparison

Which pastes best into your app?

The answer depends on the destination:

For Google Docs

Claude's cleaner output pastes slightly better because there's less formatting to process. ChatGPT works well too. Both work great with Unmarkdown™ for Google Docs.

For Word

ChatGPT's structured headings map well to Word's heading styles. Claude's output is equally good. Use Unmarkdown™ for Word for proper heading style mapping.

For Slack

Less formatting is better in Slack since it has limited support. Claude's concise style works best. All three work with Unmarkdown™ for Slack.

For Email

Any of the three work well. The conversion to inline-styled HTML handles the differences. Use Unmarkdown™ for Email.

The real answer

All three AI tools produce markdown that works well with a proper converter. The formatting differences between them are minor compared to the gap between markdown and what your destination app expects.

The tool you use to bridge that gap matters more than which AI produced the markdown. Unmarkdown™ handles output from all three (plus Copilot, Perplexity, and any other markdown-producing tool) and converts it to the format each destination needs.

Related reading

• What Is Markdown and Why Does Every AI Tool Use It?
• Why ChatGPT Output Looks Terrible When You Paste It (And How to Fix It)
• 5 Things That Break When You Paste AI Output (And How to Fix Each One)
• Claude Projects vs ChatGPT Projects vs External Knowledge Bases
• The AI Formatting Problem Nobody Talks About (And How to Fix It)
• Claude Artifacts vs ChatGPT Canvas vs Gemini Gems: Which Produces Usable Documents?