Most of the advice about spotting AI text is wrong, and it fails in both directions: it flags human writing and clears generated writing. Here is what actually holds up, written from the perspective of someone who publishes every day and has to check sources.
Why "it sounds like AI" is not a test
The signals people learn, like em dashes, tidy three-part lists, or the phrase "it is important to note", appear in plenty of human writing. They are artefacts of a style of editing, and style is copyable. Judging authorship from tone gives you a guess about how something was written, not who wrote it.
Detector tools are a different problem. They return a probability from patterns in text, and that score moves when text is edited, translated or lightly rewritten. A high score is not proof, a low score is not clearance.
The tells that survive contact with reality
- Specifics that check out. Named sources, dates, version numbers, direct quotes with a link. AI can produce these, but it produces them wrong more often. Verifying three specific claims beats any stylistic read.
- Claims that cannot be checked at all. "Studies show" with no study, "experts agree" with no expert. This is the most common pattern in low-effort generated copy.
- Uniformity at length. Long text with no digressions, no corrections, no oddly specific personal detail, and identical paragraph lengths throughout.
- Errors a human would catch. Citations to papers that do not exist, quotes attributed to the wrong person, products with features that never shipped. This is the strongest practical signal.
- Provenance. C2PA content credentials, platform AI labels, and image metadata can confirm generation. Their absence proves nothing, because most tools strip them.
How to check a text in five minutes
- Take the most checkable claim and verify it. If the piece cites a report, open the report. Two or three checks usually settle it, because fabrication clusters.
- Look at the byline and date. A named author with a history, a publication date, and a corrections policy is a stronger signal than any stylistic marker.
- Check whether the text admits uncertainty. Human writers hedge about what they do not know. Generated text tends to be confident everywhere, including about things that are wrong.
What AI labels do and do not mean
A label tells you a platform detected generation, or a creator disclosed it. Both are useful and both are incomplete. Platforms label unevenly, and creators can simply not disclose.
The practical rule: a label is evidence in one direction only. If it says AI, believe it. If there is no label, do not conclude a human wrote it.
Regulators are moving on this. The EU AI Act sets transparency duties for certain AI-generated content, and a Deloitte survey published this week found 84% of Australians want labelling to be a legal requirement, not a platform choice.
If you are the one publishing
Say what the machine did and what the person did. "Drafted with AI, checked and edited by a human" is a sentence readers can act on, and it costs nothing to write.
Keep sources visible, keep a named human accountable, and never let a tool invent a citation you did not open. That is the whole discipline, and it works better than any detector.
Originally published at voxahq.in. I write about AI tools and automation there without affiliate links in news or guides.
