What this guide is for
Spotting is the first pass. You are not proving authorship. You are looking for repeated surface patterns that justify a deeper check.
Spotting: you read and name patterns.
Checking: you collect enough text, run a tool, and interpret results with context. That workflow is covered in How to check AI writing.
The 60-second triage
If you only have a minute, scan for these three signals first:
Template language repeats
Same broad framing phrases, different sentences, little new information.
Rhythm never changes
Sentence length and cadence stay unnaturally even for full paragraphs.
Claims avoid specifics
No names, numbers, dates, or concrete examples where they should exist.
One signal is noise. Repeated signals across the same passage are a lead.
High-signal patterns
The best tells are repeated structures, not vibes. One odd sentence proves nothing. A pattern family repeating across multiple paragraphs is where confidence starts to rise.
Stock phrase templates
AI models often use polished framing phrases that sound confident but say little.
Metronomic rhythm
The prose stays on one beat: same sentence length, same cadence, same landing.
Throat-clearing intros
The draft spends the opening paragraph announcing what it will do instead of doing it.
Abstract claims without detail
Statements sound reasonable but collapse when you ask “which one, exactly?”
Pivot crutches
"Not just X, but Y." Once is fine. Several repeats in short space is a signal.
Over-signposting
"First, second, third, furthermore, finally" in rigid ladders that content does not need.
False alarms people overuse
These are weak clues that create false positives when treated as proof.
If any of these are your main evidence, pause and read the False Positive Hall of Fame first.
When suspicion becomes evidence
One flag is noise. A cluster is signal. If you can point to two or more pattern families repeating across two or more paragraphs, it's worth running the diagnostic.
How to verify without overreaching
If the evidence stays thin after all four steps, the right answer is uncertainty. That is not a failure. It is intellectual honesty.
What a score can and cannot tell you
Spotting patterns is the beginning of a process, not the end of one. If you run the diagnostic and get a number, that number reflects surface overlap with AI-typical writing — not who wrote the text. Before acting on any result, read the scoring logic and the known failure modes.