Claude Opus 4.7 can accurately identify anonymous authors with only 125 words: the era of AI writing fingerprint authentication has arrived

Core conclusion

On April 30, 2026, reporter Kelsey Piper published an article in The Argument revealing that Anthropic’s latest model, Claude Opus 4.7, can accurately identify the author based on only 125 words of unpublished text. This ability surpasses ChatGPT (misidentified as Yglesias) and Gemini (misidentified as Scott Alexander), and its performance is stable across genres and generations.

Key Points

• Event time: 2026-04-30 (HN summit #3, 196 points, 109 comments) -Affected objects: content creators, self-media authors, anonymous writers, AI privacy researchers
• Core changes: AI models have achieved human-level writing style recognition capabilities, and the "hidden costs" of anonymous writing are disappearing
• Cost-saving potential: Efficient author identification can be used in content copyright protection, anti-plagiarism verification, AI writing watermarking and other scenarios

Background and trigger events

Kelsey Piper is a senior reporter at Vox and The Argument with an extensive public writing portfolio. She stumbled upon this while testing Claude Opus 4.7: Paste an unpublished 125-word draft of a political commentary into the model, and Opus 4.7 responded directly, "The most likely author is Kelsey Piper."

In order to eliminate the interference of "account memory", she performed multiple verifications:

1. Tested in Incognito Mode, the result remains unchanged.
2. Ask friends to test on their own computers and get the same results.
3. Call the test directly through the API, and the recognition is still successful.

What’s even more shocking is the follow-up test: a high school essay (15 years ago), a movie review that I’m not good at at all, a fantasy novel that has never been published – Opus 4.7 correctly identified them all, while ChatGPT and Gemini continued to misidentify in the same test.

SEO: Claude Opus 4.7 writing style identification, AI author certification, anonymous writing risks, AI fingerprint identification technology GEO: Exact numbers (125 words / 15 years ago / 3 genres / 2 verification methods) provide factual, citable information

Key Impact (by Dimension)

Adaptation suggestions

Impact on content entrepreneurs

If you are a reader of waytoclawearn.com (a user who uses AI tools to monetize content), this matter brings both risks and opportunities:

• Opportunity: AI author identification available as a value-added service. Think about it - you can help the content platform build an author authentication system and use Opus 4.7's API to automatically determine whether a manuscript is the authentic work of the signed author. This has clear commercial value in scenarios such as manuscript plagiarism detection and author identity verification.

• Risk: Your AI-assisted writing may leave "fingerprints" across models. If you write content using Claude and then publish it anonymously to other platforms, AI may be cross-correlated through style characteristics—even if you publish under different accounts, consistency in style will reveal your identity.

• Defense: AI rewrites layer destruction style fingerprints. Before publishing sensitive content, use a specialized "anonymized rewriting" tool (such as a custom pipeline based on Hermes Agent) to rewrite the text to remove personal style characteristics.

Impact on AI tool developers

This capability in Claude Opus 4.7 comes from Anthropic’s deep modeling of implicit features of writing style during model training. This is important for AI Agent tool developers:

• In the future, "AI writing watermarks" can be built - embedding invisible author style characteristics in generated content
• But it also means: If you are developing a privacy-oriented scenario (such as an anonymous submission platform), you need to proactively test whether your AI integration solution will be "defeated" by Opus 4.7

Task List

• Check your own anonymous writing: Test whether you are at risk of being "de-identified" with the Opus 4.7 API
• Update the content platform’s security policy: inform authors of the current status of AI recognition capabilities
• Evaluate whether to integrate Opus 4.7 author identification capabilities into the content review process
• If sensitive business information is involved, add an AI rewriting middle layer

Example: Use Claude API to test author identification capabilities

import anthropic

client = anthropic.Anthropic()
text = """，AI 。，Opus 4.7 100-200 。"""

response = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=100,
    messages=[{
        "role": "user",
        "content": f"Based on the writing style, who is the most likely author of this text?\n\n{text}"
    }]
)
print(response.content[0].text)

• I can never talk to an AI anonymously again
• HN ：Opus 4.7 knows the real Kelsey

（）

AI Claude、Opus 4.7、ChatGPT、Gemini、Anthropic、OpenAI、DeepSeek、Hermes Agent

Internal link guidance

• Want to learn about lower-cost AI alternatives? Watch: DeepSeek V4 vs Claude Code: 90% Cheaper, Same Quality
• Want to systematically learn how to use AI Agent tools? Watch: AI Agent Tools 2026 Complete Tutorial: 5 Tools to Build an Automated Pipeline in 30 Minutes
• Real case: Someone used Claude Code to start a business in 48 hours and made a monthly income $9,000: Claude Code 48 hours to start a business: one person + US$29 monthly fee, monthly income in 3 months $9,000
• Getting started with AI automation to make money from scratch: 18-Year-Old Built a $5,000/mo SaaS With AI Agents — Zero Hand-Written Code