Winston AI

@Winston AI can you check if this draft blog post was written by a human or AI? [paste 500-word article]

Analysis complete. The text shows a 78% probability of AI generation, with high confidence in paragraphs 2-4 (92-95% AI likelihood) and mixed signals in the introduction (45% human-like). Key indicators: repetitive sentence structures, lack of personal anecdotes, and uniform tone throughout. The conclusion paragraph appears more human-authored (32% AI probability). Overall assessment: likely AI-generated with possible human editing in specific sections.

This example demonstrates Winston AI's core detection capability. The MCP returns probability scores, not binary judgments — useful for content review workflows. Note that detection accuracy depends on text length (minimum ~50 words recommended) and the AI model that generated it. Results are probabilistic assessments, not definitive proof.

@Winston AI I need to verify this student essay hasn't been copied from online sources. Can you run a plagiarism check? [paste 800-word essay]

Plagiarism scan completed against online databases. Found 3 matches: 12% similarity with a Wikipedia article on climate change (paragraphs 1 and 3), 8% match with an academic journal abstract from 2021, and 5% overlap with a blog post from 2023. Total plagiarism score: 25%. The remaining 75% shows no direct matches in indexed sources. Detailed match locations and source URLs provided for review.

Plagiarism detection requires the text to be 'prepared' first (a separate step in the MCP workflow). This example shows how Winston AI compares submitted text against its indexed web corpus. The tool identifies percentage matches and source links, but cannot detect paraphrased content or offline sources. Useful for academic integrity checks and content originality verification before publication.

@Winston AI compare these two product descriptions and tell me if one is too similar to the competitor's version. [paste two 200-word descriptions]

Text comparison complete. Similarity score: 64%. Both descriptions share identical phrasing in 8 sentences, particularly around feature lists and technical specifications. Structural overlap is high — both follow the same three-paragraph format with benefits-first positioning. Unique elements: your version includes customer testimonial language (15% of content), while the competitor's focuses on pricing tiers (12% of their content). Recommendation: 64% similarity may raise concerns in competitive contexts; consider rephrasing shared feature descriptions to reduce overlap below 40%.

The TEXT_COMPARE tool measures textual similarity between two documents — helpful when assessing whether marketing copy, documentation, or product descriptions are too derivative. This example pairs the comparison with AI reasoning to interpret the score and suggest next steps. Note that high similarity isn't always problematic (industry-standard terminology is expected), so context matters when acting on results.

When Winston AI catches copy-paste before publication

A 6-person content team ships 40 blog posts a month across three writers. The editor suspects one contractor is recycling competitor content. Winston AI's plagiarism detection tool runs each draft against online sources before the final review. The MCP needs only an API key and exposes three focused tools—no sprawl, no setup friction. If a post flags above 15% similarity, the editor sends it back with the matched URLs. This works when you're auditing finished drafts, not real-time collaboration. If your team writes in Google Docs and needs inline feedback, Winston AI won't integrate there—you'll copy-paste into Switchy. For monthly audits on a small team, that trade-off is fine. Add Winston AI to your Switchy workspace if you're publishing contractor work and need a second opinion before it goes live.

Why Winston AI fits adjunct faculty grading 80 essays

An adjunct professor grades 80 essays every two weeks across three sections. She suspects 10-15% are ChatGPT rewrites but doesn't have time to manually vet each one. Winston AI's AI text detection tool scores each submission in under a minute. She pastes the essay into Switchy, runs the detection, and flags anything above 70% AI probability for a closer read. The MCP's three tools cover her workflow: detect AI, check plagiarism, compare two versions if a student revises. This breaks down if you're grading 500 essays—Winston AI charges per scan, and the cost stacks fast. It also won't catch sophisticated prompt engineering or heavy human editing. For an adjunct workload where you're triaging the obvious cases, Winston AI gives you a fast filter. Connect it to Switchy if you're grading under 100 submissions per cycle and need a first-pass screen.

When Winston AI verifies associate drafts before filing

A 3-attorney firm files 20 motions a month. The senior partner reviews every brief an associate drafts, but she's burned twice by associates lifting language from old filings without attribution. Winston AI's text comparison tool measures overlap between the new draft and the firm's archive. She runs the check in Switchy before the final proofread. If similarity exceeds 40%, she asks the associate to rewrite or cite. The plagiarism detection tool also scans public case law to catch uncredited quotes. This works because legal writing has a narrow scope—you're comparing against a known corpus, not the entire internet. If your firm handles 100+ filings a month, you'll want a document management system with built-in version control instead. For a small practice where the partner is the bottleneck, Winston AI adds a 5-minute safety check. Use it in Switchy if you're reviewing associate work and need to verify originality before the brief leaves the office.