AI Detector
Free AI Check

AI Detector

This AI detector reads a piece of text and reports how much of it looks machine-written. The result is a detection score plus a marked version of your text, split into AI-generated, human-written, and mixed passages.

  • Paste text or upload a file
  • Up to 2000 words per check
  • A score with the text marked by category
  • Covers output from several writing models
  • No account needed to run a check
Detect AI

What Is an AI Detector

An AI detector reads text and reports how likely it is that a machine wrote it, marking the passages it reads differently.

An AI detector takes a piece of writing, analyses its sentence patterns and word choices, and returns a score with the text marked up. The marking sorts the passage into three groups: text that looks AI-generated, text that looks human-written, and text that sits between the two. What the tool produces is a likelihood, not a verdict. Two pieces of writing can score differently for reasons that have nothing to do with who wrote them, which is why the score is worth reading together with the marked sections rather than on its own.

Why Choose This AI Detector

The report is built to be read: a score, the passages it came from, and a second look at anything that is unclear.

Paste the text or upload a file

Text can be typed or pasted into the input area, or uploaded as a file when the writing already lives in a document. One run covers up to 2000 words.

Read a single detection score

The result opens with one number describing how likely the text is to be machine-written. A score at either end of the range is easier to read than one in the middle, and the middle is where most texts land.

See which passages were flagged

The text comes back marked, with AI-generated, human-written, and mixed passages shown in different colours. Reading those sections is what explains the score.

Check text from the common writing models

The analysis covers output from tools such as ChatGPT, Gemini, Copilot, and Claude. A passage produced by one model and edited in another still gets checked as one piece of text.

Run the check again on an edited version

Check Again sends the current text back through the analysis, and Clear empties the field for the next document. Running a draft before and after editing shows what changed and what did not.

Hand a flagged text to a rewriting step

When a result comes back largely machine-written, the same screen offers a route to a humanizing tool. That keeps detection and rewriting in one place rather than across two tabs and two pastes.

AI Detector

AI Detector

@tusummary

Pain point: a submission looks machine-written and there is no way to say why

Editors and teachers read a lot of text and notice when something is off before they can name it. The paragraphs are even, every claim arrives in the same shape, and none of the sentences take a risk, but none of that is evidence. The AI detector puts a number and a set of marked passages next to that impression. It does not settle the question, but it gives the conversation something concrete to look at.
Check a Text
Draft submission open with paragraphs marked
AI Detector

AI Detector

@tusummary

Pain point: the result says mixed and nobody knows what that means

A mixed reading is the most common outcome and the least useful one on its own. Text generated by a model and then rewritten by a person carries both sets of patterns, and the tool reports what it finds rather than where the writing came from. Reading the marked sections is what turns the label into information. If the flagged passages are the parts that were generated first and then edited heavily, the result matches what the writer already knew about how the text was made.
Check a Text
Text shown with three different highlight colours
AI Detector

AI Detector

@tusummary

Pain point: the text was generated and then rewritten by hand

The usual workflow is not one or the other. A draft comes out of a model, gets cut down, reordered, and rewritten in the writer's own voice, and the finished piece is genuinely neither of the two things a simple check is looking for. Running the check on the rewritten version shows what is left of the original pattern. Comparing that with an earlier run on the raw draft shows what the editing changed.
Check a Text
Generated paragraph with heavy hand edits
AI Detector

AI Detector

@tusummary

Pain point: a policy exists and every case is judged differently

A newsroom, a department, or a publisher decides to screen submissions, and the first month goes by feel. One person lets a piece through and another holds back something similar, and the inconsistency becomes the story. A check that returns the same kind of report each time gives everyone the same starting point. The score, the marked passages, and the follow-up conversation have the same shape in every case, which is what makes a policy possible to apply.
Check a Text
Policy document beside a set of detection reports
AI Detector

AI Detector

@tusummary

Pain point: the text was translated and the origin is unclear

Text that passed through a translation step reads unlike either language it came from. Sentences that were normal in the source arrive long and evenly weighted, and a check run on the result can react to that shape rather than to who wrote the original. Knowing that the translation happened is part of reading the score, because the flagged passages often line up with the sections that came through the translation first.
Check a Text
Same article in two languages on a desk
AI Detector

AI Detector

@tusummary

Pain point: the writer knows the words are their own

Some writing scores badly for reasons that have nothing to do with how it was produced. A tight, formal register, a list of parallel clauses, and a habit of writing to a template all look like machine output when the text is read by a model. Running your own draft through the check shows whether that is happening. The marked passages point at the sentences worth loosening, which is a style note whether or not the score is fair.
Check a Text
Writer reading a score on their own draft
AI Detector

AI Detector

@tusummary

Pain point: the document is far longer than one check allows

Reports, theses, and long features do not fit into a single pass. Splitting a document at the wrong place hides the pattern, because the sections that read the same way often sit next to each other. At 2000 words per run, a long document can be worked through in chapter-sized pieces. Checking each piece against the previous result shows whether the flagged passages cluster in one part of the text or run through the whole thing.
Check a Text
Long report split into sections for checking
AI Detector

AI Detector

@tusummary

Pain point: a student and a teacher need to talk about the same text

The conversation about a suspicious assignment goes badly when it starts with an accusation. Nobody in the room wrote the analysis on the screen, and the writing itself is the only shared object to point at. Opening the marked report turns the discussion toward specific passages and specific sentences. It reframes the topic from character to text, which is a discussion both sides can take part in.
Check a Text
Two people reading the same report together
AI Detector

AI Detector

@tusummary

Pain point: the client wants reassurance before publishing

Freelancers get asked to confirm that a piece is original, and a denial carries no weight without something to show. The question underneath is usually about risk, not about authorship. A detection report gives the exchange something concrete to attach to, and both sides can then decide what to do about the passages that came back flagged.
Check a Text
Report attached to a client message thread

How to Use This AI Detector

Four steps, and the third one is where the score gets read rather than taken at face value.

Step 1

Add the text

Paste the passage into the input area or use the file option to bring in a document. The counter shows how much of the 2000-word limit the text is using.

Step 2

Run the check

Press Detect AI. The analysis runs across the whole passage and returns a report rather than a single yes or no.

Step 3

Read the score with the marked text

Look at which passages were flagged and what they have in common. The categories explain the number, and the number on its own rarely tells you what to do next.

Step 4

Act on the result

Use Check Again on an edited version, clear the field for the next document, or move a heavily flagged text into the humanizing step offered on the same screen.

Who Uses an AI Detector

It fits anywhere a text arrives from outside and someone has to form a view of it.

ES

Editors screening work

Unsolicited submissions and pitches

A submission that reads too smoothly is hard to question without sounding paranoid. A marked report gives the editor a specific set of passages to raise, and the writer a specific set to answer for.

TR

Teachers reviewing work

Assignments and coursework

The report is a way into the conversation rather than the end of it. Discussing two flagged paragraphs is a different meeting from accusing a student of something the tool cannot prove.

CC

Clients checking copy

Freelance articles and marketing text

A buyer wants to know what they are publishing. Seeing the score and the marked passages before the piece goes live is usually enough to settle the question and keep the schedule.

Frequently Asked Questions

How the check works, what the score means, and the limits worth knowing before you rely on a result.

Read the text before you judge it

Add up to 2000 words, run the check, and look at the marked passages behind the score.

Check a Text
A Signal, Not a Verdict

The AI detector reports likelihood, and likelihood can be wrong

There is no reliable marker inside machine-written text, so every detection result is an estimate. Formal registers, translated writing, and heavily templated prose can score as AI-generated when a person wrote them, and edited machine text can score as human. Use the report to decide what to look at closely, never as the only reason to reject a piece of writing or to question the person who submitted it.