Is this email or text message AI?
Paste an email or a text message to get an estimate of how likely it was written by AI. Free, and it runs in your browser.
Is this email AI-generated?
Remove names, numbers and codes before pasting. Your text stays in your browser.
—
Sentences that read most like AI
This is an estimate, not proof. AI detectors make mistakes, especially on short, formal, or templated messages, and on writing by people using a second language. Don't accuse anyone, or decide someone is a scammer, based on this score alone.
Can you tell if an email is AI-generated?
Sometimes, but never with certainty. AI chatbots are now used to write sales emails, cover notes, customer-service replies, and scam messages. Their writing often has a recognizable style: even sentence lengths, a polished and formal tone, and few contractions. Detectors look for these patterns. The problem is that plenty of people write that way too, especially in work email, and many companies use templates that read just like AI.
This tool runs the same checker as our main AI detector: two open-source AI-detection models (based on RoBERTa) running in your browser, plus a small weight from writing-pattern checks adapted from Wikipedia's "Signs of AI writing" guide. You get a percentage and a plain label: under 30% is likely human, 30 to 80% is uncertain, and over 80% is likely AI.
How the email check works
Pasted emails often include extra text that isn't part of the message. Before scoring, the checker removes header lines (From:, To:, Subject:, Date:), lines quoted from earlier replies (starting with ">" or after "On ... wrote:"), and anything after a signature marker like "--" or "Sent from my iPhone." That way the score reflects the message itself, not a forwarded thread or a company footer.
Checking text messages: why short is hard
Most texts are a sentence or two. Detectors need enough words to spot a pattern, and a short message just doesn't have much to go on. "Running late, be there in 10" looks the same whether a person or a phone's suggested reply wrote it. So we're upfront about it: messages under about 12 words aren't scored at all, and anything under 50 words gets a "low confidence" label. Longer texts, like a scam message about a package or a bank account, give the checker more to work with, but the result is still only an estimate.
What to do with the result
- Use it as one clue. A high score means the writing looks like typical AI output. It doesn't tell you who sent it or whether it's honest.
- For suspicious messages, focus on the request. Urgent requests for money, gift cards, login codes, or personal details are red flags no matter who wrote the message. Contact the sender through a number or website you already trust.
- Be fair to people. Don't accuse a coworker, applicant, or friend based on a score. Detectors flag human writing more often than you'd expect, and research has found they flag non-native English writers far more often (Liang et al., 2023).
- Look at the wording yourself. Our AI Words Finder highlights the words and phrases that show up a lot in AI writing.
Related guides
- Are AI detectors accurate?
- Signs text was written by AI
- Why AI detectors flag human writing
- AI Words Finder
FAQ
Can you tell if an email was written by AI?
Not for certain. This checker gives an estimate based on two AI-detection models and some writing-pattern checks. Polished business emails and templates can look AI-like even when a person wrote them, so treat the result as a hint, not proof.
Can it check a text message?
Yes, but be careful with the result. Most text messages are only a few words long, and very short messages can't be judged reliably by any detector. The page warns you when a message is too short for a confident result.
Is the text I paste sent anywhere?
No. The check runs in your browser. Your text isn't sent to our server or stored. Your browser downloads the detection models once (from Hugging Face and our site) and then reuses them. Even so, remove names, phone numbers and codes before pasting.
Does it check the whole email, including the signature?
It tries to check only the message itself. Before scoring, it drops header lines like From: and Subject:, quoted earlier replies, and anything after a signature line ("--" or "Sent from my phone").
Can it tell which AI wrote the message?
No. No detector can reliably tell which AI tool wrote a text, so we don't try. The result is only an estimate of how AI-like the writing is.
Is a phishing email always AI-written?
No, and an AI-written email isn't always a scam. Whether a message is safe depends on who sent it and what it asks you to do, not on who or what wrote it. If a message asks for money, a code, or a login, check with the sender another way.
Sources
- Wikipedia contributors (WikiProject AI Cleanup), "Wikipedia:Signs of AI writing." https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing (CC BY-SA 4.0)
- Liang, Yuksekgonul, Mao, Wu & Zou, "GPT detectors are biased against non-native English writers," Patterns 4(7), 2023. https://arxiv.org/abs/2304.02819