How to tell if a text was written by AI

A sign is not proof

Read the whole text once, from the first line to the last, before you search it for any single sign of AI writing. This comes first because a sign only means something next to the rest of the text. A phrase that seems strange on its own may suit the topic, and a smooth paragraph may be the place where the writer worked hardest. If you want to know how to tell if an essay was written by AI, this first reading is where a fair answer starts. After it, you can look for the signs below and keep the ones that help apart from the ones that mislead. Reading can give you a fair reason to doubt a text, but it cannot give you proof, and a person still has to decide what happens next.

Signs that help, together and not alone

One sign says very little. In our view, what deserves attention is several signs in the same text at once. The most used AI writing tools have habits that show when nobody edits their text, and those habits show more in how the text is built than in single words. So look at the shape of the whole piece before you study its vocabulary.

Shape gives the first signs. Raw AI text, meaning text a tool produced that nobody has edited, often has balanced sections of the same size even when the topic does not ask for it. Compare the length of the sections, and ask yourself whether the topic really calls for parts of equal size.

The content gives more. Watch for confident, correct sentences that commit to nothing: no names, no examples of the writer’s own, no detail that only this writer could know. Then check the sources. Look up a few of the quotes, books and articles the text relies on. A quote that cannot be found, a book that does not exist, or a real source that does not say what the text claims tells you more than any choice of words. Last, read the task again. A text that smoothly answers a similar question, or ignores what the class worked on, deserves a second look.

If you know the writer, put the text next to their earlier work. A sudden change in the words, in the level of detail or in the way sentences are built is a sign to keep in mind, though people do improve, and a new course can change how someone writes.

One sign is fairly clear: leftover parts of a chat. These are lines addressed to the person who asked for the text, such as an offer to make it longer or a remark about what the chat tool cannot do. If you find one, talk to the writer first, and decide nothing until you have.

Careful, correct writing is not a sign of AI

Some features are often read as signs of AI and tell you nothing about it. Start with good spelling. Schools teach students to write without mistakes, many readers expect it, and plenty of writers check their work slowly or use a spelling checker. If you suspect a text because it is well written, you end up accusing the people who tried hardest.

A formal tone tells you just as little, because it is the tone taught for school and university work, and many work texts use it too. The plan of an opening, a body and an ending is no different. Students practise that plan for years. Many tasks ask for it, and a writer who follows it is doing the task well.

People who write in a second language often use steady patterns that they learned at school. A reader can mistake that steadiness for AI, yet it is only one way of writing English. Words and sentence forms from one region are part of the language too, and they give no reason for doubt.

A long dash says nothing about how a text was made. Many writing programs use it, and so do many careful writers. Grammar checkers and translators cause more confusion. A text that was run through one can resemble AI text even when the ideas are the writer’s own, and our guide on translation tools and AI checkers explains why.

Why no sign proves anything

Every sign above describes a way of writing. None of them describes the person. Some people write even paragraphs and careful general sentences with no help at all. The habits also change each time a tool is updated, so no list of signs stays right for long. No one can know for sure which tool produced a text, so naming the tool is not a useful aim.

Reading gives you a fair reason to doubt, and nothing more. Better information comes from how the text came about: the drafts, the version history of the document where one was kept, and the sources the writer used. Someone who did the work can usually explain it in their own words.

What to do if you suspect a text

A suspicion is where a conversation begins. If you are a teacher, speak to the student in private before you decide anything. Show the passages that made you doubt and ask how the work was done. You could ask why the student chose a certain source, or what changed between the first draft and the last. Hear the answer to the end before you form a view.

If you edit texts or commission writing for work, ask for the drafts and the list of sources, then check the references yourself. Quote to the writer the exact lines that raised your doubt. Reject the piece only after they have had a chance to answer.

If you are a student and worry that your own work may raise doubt, keep everything that shows how you worked: your drafts, your outline, your sources and the version history of the document. Do not change those files once the work is handed in. Then you can show each step if anyone asks.

What an AI checker can tell you

Think of an AI checker’s result as an estimate. The checker makes it from signs in the writing, and it can help you decide where to look. Still, the estimate can be wrong in both directions. A text that a person wrote is sometimes marked, and a text that an AI tool produced is sometimes missed. Your own reading and the checker’s result are both only reasons to look more closely. The result is a signal. A person has to weigh it.

After you have read a text yourself, you can scan it with the AI checker on ScanForAI and compare the result with what you noticed. Neither your reading nor the result proves who wrote it. Where the two agree, you know which passages to ask the writer about. Where they differ, read those parts again and check whether one of the misleading signs explains the difference.

Some AI tools can leave a hidden mark in their text, but finding no such mark proves nothing either way.

If other hands or tools have changed a text, the signs in this guide tell you much less. A text that has no AI in it can pick up these habits from a translator or a grammar checker. And the habits of the AI tools themselves change each time a tool is updated. Set the signs aside in that case and go by how the text was made: ask for the drafts and the version history, and talk with the writer about the work.

This guide is general information for students and teachers. It is not legal advice, and your school’s own rules come first. A score from any AI checker is a signal, never proof.

Cite this guide

ScanForAI team (2026). How to tell if a text was written by AI. ScanForAI. https://scanforai.com/guides/can-you-spot-ai-writing-by-reading-it

You may share and adapt this guide if you credit ScanForAI and link to it. See Reuse and citation.

AI detection is not 100% accurate. No detector, including ScanForAI, can prove if a text was written by a person or by AI. Every result is an estimate based on writing patterns. It can be wrong in both directions. Never use it as the only reason for an important decision about a person, such as a school decision, a punishment or a hiring choice. Never use it to accuse someone.