How can you tell if content was written by AI?
Why AI detectors are not the answer
AI detectors regularly flag genuine writing as AI, especially formal writing and writing by non-native speakers, and they miss AI text that someone has edited. Google itself does not judge content by who or what wrote it. It judges whether the content is helpful, and it asks publishers to fact-check anything generated by AI before publishing it.
The signs worth noticing
- Nothing first-hand: no "I tested", no real examples, dates, photos or numbers that only someone who did the work would know.
- Stock phrases: "delve", "in today's fast-paced world", "it is important to note", "leverage", "crucial", "seamless".
- Long dashes (—) used far more often than most people type them.
- Every section built the same way, with sentences of very similar length.
- Statistics and "studies show" with no source anywhere near them.
- No real, accountable author, or an author with an AI-generated headshot, which Google now describes as deception.
None of these proves anything on its own. Professional writers use long dashes too. It is when several appear together that they tell you something.
The test that always works
I build audit tools, and the most important thing I have learned is to check every finding against the actual rule rather than trust the tool. The same works for content. Pick a specific claim, such as a statistic, a date or a rule, find the original source it should come from, and see whether the source really says it. Content written by someone who knows the subject holds up. Lazy AI content often fails on the first check.
Our free E-E-A-T Checker now reports these writing signals with the evidence for each one, and lists any statistics that have no source nearby so you can check them yourself. It never labels a page as "AI-written", because no tool can prove that.
Sources
- Google Search Central: Creating helpful, reliable, people-first content
- Google Search Central: Guidance on using generative AI content on your website
Frequently asked questions
How can you tell if content was written by AI?
often you cannot prove it, and AI detectors are unreliable. What you can spot is lazy AI content: generic wording, no first-hand detail, stock phrases, long dashes everywhere, and confident claims that turn out to be wrong. The one test that always works is to check the claims against their original sources.