ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›RECORD›HOW TO TELL IF SOMETHING IS WRITTEN BY AI: 9 SIGNS (AND WHY DETECTORS FAIL)
RECORD · REVISED ON A SCHEDULEBEGINNER

How to Tell if Something Is Written by AI

How to tell if something is written by AI: nine signs to look for, why none of them proves it, why detectors fail, and what to check instead.

READ9 min
WORDS1,737
SECTIONS5
SOURCES1
TYPEREVISED
CHECKED25 AUG 26
TL;DR — THE SHORT VERSION

Nine signals can shift your confidence that text is machine-written, none of them proves it, detector tools are too unreliable to accuse anyone with, and whether a text is true matters more than who wrote it.

  • Signals, not proof. Details that do no work, balance with no position, even rhythm and empty transitions raise suspicion, but each has a human explanation too.
  • Detectors misfire. Benchmark research found detectors struggle with models and topics they were not tested on, and simple rewording cuts their accuracy sharply.
  • Careful non-native writers get flagged. A 2023 Stanford study found detectors wrongly flagged a large share of essays by non-native English speakers, while doing near-perfectly on US eighth-grade essays.
  • Ask about process instead. Drafts, sources and the decisions behind a piece are fairer evidence than a score.
  • Check claims, not authors. A machine-written guide with correct facts serves you better than a human-written one that is wrong.

You can sometimes tell if something is written by AI from nine signs, such as details that go nowhere, balance with no opinion, and an even, same-y rhythm. None of them proves it, and AI detector tools are not reliable enough to accuse anyone. The fairest test is to check whether the claims are true, and to ask the writer how they made it.

◈ READ THIS FIRST

Nothing below proves anything. These are signals, and signals shift your confidence rather than settle a question. The moment a signal becomes an accusation, it is being used wrongly — including by software that claims to do it for you. Machine text that a person edited and careful human text look increasingly alike, and that convergence is permanent.

The nine signs of AI writing

Sign 1: Specificity that goes nowhere

Details appear and then do no work. A year is named but nothing depends on it; an example is offered but the argument would be identical without it. Models produce the texture of expertise because texture is what they learned. Writing that actually knows a subject tends to let its details change the conclusion.

Weak when: the writer is padding to a word count. Humans do this too.

Sign 2: Perfect balance, no position

Every consideration weighed, every side acknowledged, nothing risked. "It depends on your needs." "Both approaches have merit." A model trained to be inoffensive drifts to the midpoint of everything it has read. People who have actually done the thing usually have an opinion about it.

Weak when: the genre demands neutrality — encyclopedias, official statements, and this site's own stance boxes.

Sign 3: Rhythmic sameness

Sentences of similar length, paragraphs of similar shape, each section built like the one before. Human writing lurches: a long thought, then three words. Machine writing breathes evenly, all the way down.

Weak when: the writer is trained to a house style, or the text is short.

Sign 4: The triple

Three items where two or four would do — clear, concise, and compelling. Three is the pattern language settles into when nothing is pushing back, and models learned it from a mountain of marketing prose.

Weak when: three is genuinely the number. Rhetoric loves threes for good reasons.

Sign 5: Transitions doing no work

"Moreover." "Furthermore." "It is worth noting that." Connectives that announce a relationship the sentences do not actually have. A person adds a transition when the logic needs one; a model adds one because paragraphs usually start with them.

Weak when: the writer learned formal academic English, where these are conventional.

Sign 6: No memory, no cost

Nothing in the text was learned the hard way. No dead end, no revision, no "I assumed X and was wrong." Real expertise carries scar tissue — the thing that was tried, failed, and changed the writer's mind. Machine text has no history to draw on, so it presents knowledge as if it had always been obvious.

Weak when: the format excludes personal experience — reference pages, product copy.

Sign 7: Summary that adds nothing

A closing paragraph restating the piece without advancing it: "In conclusion, X remains a complex and evolving space." Models are strongly shaped toward wrapping up. Writers who are finished tend to simply stop.

Weak when: the piece is long enough that a real summary helps.

Sign 8: Confidence uncoupled from evidence

Assertions delivered in the same even tone whether they are well established, contested, or invented. Fabricated citations sit beside real ones in identical typography. A model has no internal sense of which of its claims are shaky, so nothing in the prose signals it.

Weak when: the writer is simply overconfident. Also common in humans.

Sign 9: No one in the room

The hardest to describe and the most reliable in practice. No particular reader is imagined, no relationship assumed, no particular person addressed. The text is aimed at everyone, which means it lands on no one. This is what people are reacting to when they say a page feels hollow without being able to say why.

Weak when: you are tired or predisposed. This signal is the easiest to hallucinate.
THE SCAN
Nine signals, one pass. None of them proves anything on its own.
TAKEAWAY

Use the signals to decide what deserves a closer look, not to decide who wrote something: each of them also has an ordinary human explanation.

Words and phrases AI overuses

The plain point: some words became far more common after ChatGPT arrived, which tells you about writing in general, not about any one text.

The best-known measurement looked at over 15 million biomedical research abstracts from 2010 to 2024. After chatbots arrived, certain style words jumped in frequency: rarer ones such as delves, underscores and showcasing, and common ones such as potential, findings and crucial. From that jump the researchers estimate that at least 13.5% of 2024 abstracts were processed with AI.1

Two limits. The study measured science abstracts, so its word list may not match emails, essays or social posts. And it counts words across millions of texts: one delve in one essay proves nothing, because people used all of these words before chatbots did.

TAKEAWAY

A pet chatbot word is a reason to look closer, never a verdict on its own.

Why AI detectors fail

The plain point: AI detector tools guess, and they guess wrong often enough that nobody should be accused on a score.

The RAID benchmark (ACL 2024), whose paper notes that commercial detectors routinely claim 99%+ accuracy, tested twelve detectors on over 6 million machine-written texts and found they struggle with models and topics they were not tested on, and that simple rewording cuts their accuracy sharply.2 A 2023 Stanford study found seven detectors wrongly flagged an average of 61.3% of essays by non-native English speakers, while doing near-perfectly on US eighth-grade essays. When the same essays’ vocabulary was polished with ChatGPT, the false-positive rate fell to 11.6%: same writers, same ideas, fancier words.3 In other words, a detector can end up penalising people who write English carefully because it is their second language. Do AI detectors work? sets out the full evidence.

DETECTOR FALSE POSITIVES ON NON-NATIVE ESSAYS
Average across seven detectors, on the same essays before and after their vocabulary was enhanced. The same detectors were near-perfect on US eighth-grade essays.
Original essays61.3%
Same essays, vocabulary enhanced11.6%
Liang et al., Patterns 4(7), July 2023, read at source 17 Sep 2026: “the average false-positive rate dropping by 49.7% (from 61.3% to 11.6%)”.
TAKEAWAY

Do not treat a detector score as evidence against a person. Research on these tools finds them unreliable, and people who write English carefully as a second language are among those they wrongly flag.

The habit that beats all nine

Stop asking who wrote this and start asking is this true, and does it help me. Authorship is nearly unprovable and getting more so; usefulness is testable today. A machine-written guide with correct, checkable facts serves you better than a human-written one that is wrong. The reason AI text is worth noticing at all is that it is frequently confident and wrong — so verification, not attribution, is the skill that pays.

For images and video rather than text, the companion guide is How to spot AI-generated images, video and voices.

TAKEAWAY

When a text makes you suspicious, check its claims before worrying about its author: accuracy can be tested today, and authorship mostly cannot.

◈ WHERE THIS SITE STANDS

AI-written text is not automatically bad and using it is not automatically dishonest — this site's own use of these tools is disclosed on its About page. What matters is disclosure and accuracy. The harm is not that a machine wrote something; it is that unchecked machine text spreads confident errors at volume, and that people are being accused of using it on evidence too weak to support the charge. Both things are worth being angry about, and they point in the same direction: check claims, not authors.

Questions people ask

Can teachers tell if you used ChatGPT?

Not reliably from the text alone. The signs on this page are hints, and detector tools misfire, especially on careful writers whose first language is not English (see why AI detectors fail). What a teacher, editor or employer can fairly do is ask about process: earlier drafts, sources consulted, why this example rather than another, what got cut. Someone who wrote the work can talk about those decisions for as long as you like, and a paraphrasing tool cannot fake that conversation.

Is AI-assisted the same as AI-written?

No. Using AI to check spelling or suggest ideas is different from handing in text a machine wrote, and schools and employers draw the line in different places, so check the rule that applies to you. In the UK, for example, the exam boards’ body JCQ says AI-generated content must be referenced with “the name of the AI tool used”.4 None of the nine signs can tell light AI help from full AI writing.

Can AI text be watermarked?

Some of it is. Google’s SynthID can hide a watermark in AI-made text, and Anthropic says text generated by Claude carries embedded watermarks, but heavy rewriting or translation weakens a text watermark, and each company’s mark is read with its own key.567 So a missing watermark does not prove a person wrote something. AI watermarking explains how it works.

Related on this site: How to Proofread Your Writing With AI · Before You Publish AI Content

SOURCES
  1. 1Kobak et al., Delving into LLM-assisted writing in biomedical publications through excess vocabulary, arXiv:2406.07016, first posted 11 Jun 2024, revised 3 Jul 2025, published in Science Advances 11(27), 2 Jul 2025, read at source 5 Oct 2026“we study vocabulary changes in over 15 million biomedical abstracts from 2010--2024 indexed by PubMed”; “This excess word analysis suggests that at least 13.5% of 2024 abstracts were processed with LLMs.” From the full text: “Less common words with strong excess usage included delves”, “underscores” and “showcasing”; “More common words with strong excess usage included potential”, “findings” and “crucial”. ↑
  2. 2Sources: Dugan et al., “RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors”, read at source 11 Sep 2026 — “over 6 million generations spanning 11 models, 8 domains, 11 adversarial attacks and 4 decoding strategies”. ACL 2024 (6.2m generations, 11 models, 8 domains, 11 adversarial attacks; 12 detectors evaluated) · Liang, Yuksekgonul, Mao, Wu & Zou, “GPT detectors are biased against non-native English writers”, Patterns 4(7):100779, July 2023.↑
  3. 3Liang et al., Patterns 4(7), July 2023, read at source 17 Sep 2026“the average false-positive rate dropping by 49.7% (from 61.3% to 11.6%)”, against “near-perfect accuracy for US eighth-grade essays”. ↑
  4. 4JCQ, Information for candidates: Coursework assessments, effective from 1 Sep 2026, read at source 12 Nov 2026 (replaces the JCQ AI guidance page this line cited until 12 Nov 2026, which no longer loads) (as cited on this site’s AI for students page): “Where AI has been used, your reference must show the name of the AI tool used and should show the date the content was generated.”↑
  5. 5Google AI for Developers, SynthID, last updated 9 Apr 2025, read at source 1 Oct 2026 (as cited on this site’s watermarking page): SynthID “watermarks and identifies AI-generated content by embedding digital watermarks directly into AI-generated images, audio, text or video.” and “Detector confidence scores can be greatly reduced when an AI-generated text is thoroughly rewritten, or translated to another language.”↑
  6. 6Claude Help Center, How Claude marks AI-generated content, read at source 13 Oct 2026“Generated text will carry embedded watermarks”. ↑
  7. 7Anthropic, How Claude’s text watermark works, 14 Aug 2026, read at source 22 Sep 2026 (as cited on this site’s provenance page), on another company’s watermark: “it would have a different key; it might also use a different watermarking method altogether”.↑

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
◈ IF YOU ARE CITING THIS

Or check it yourself. How to check the figures here names the feed or document behind each recurring source, and what to expect when your number differs from ours.

Cite the original source, not this page. Every figure here names the organisation that issued it and the date it was published — those are the citations worth carrying. This page is a signpost, not a primary source.

If you need to reference the collation itself — the comparison, the framing, or a correction made here — the press page has the details. But if you are quoting a number, go to whoever measured it.