Before you publish anything in 2026, ask one question: could a generic ChatGPT prompt have produced something very similar to this? If the honest answer is yes, the piece isn't ready, it's missing an opinion, a data point, a specific experience, or a contrarian angle that only you could have added.
This isn't a philosophical exercise. It's a five-minute pre-publish check that catches content before it goes live and quietly underperforms because it's indistinguishable from what a thousand other sites already generated.
Key Takeaways
- The core test: describe your article's topic in one prompt-length sentence, imagine the generic AI output, and compare it honestly to your draft.
- If your draft and the imagined AI output are substantially the same, the piece is missing an originality ingredient, not a length or formatting problem.
- The four ingredients that reliably beat generic AI output: a real opinion, original data, a specific personal or client experience, and a contrarian but defensible angle.
- Score each draft on a simple 0-4 rubric (one point per originality ingredient present) before publishing.
- A piece scoring 0 or 1 should be rewritten or killed, not published as-is, publishing it does more harm than not publishing at all.
- This test works because AI Overviews and chat assistants already know generic content; they cite what they don't already know.
Why This Test Exists
Every SEO team now has the same problem: AI writing tools made it trivially easy to produce "complete" content, and that ease flooded the web with pages that are structurally fine and substantively empty. Search Engine Journal and Search Engine Land have both covered how Google's ranking systems and spam policies increasingly target this exact pattern, content that's well-formatted, keyword-covered, and functionally interchangeable with a dozen competitors.
The problem isn't that AI was used to help write something. It's that too much published content, AI-assisted or not, contains zero information the reader (or the AI system reading on the reader's behalf) didn't already have access to. Google's own guidance on creating helpful, reliable content is explicit that the test isn't how content was produced, but whether it demonstrates real experience and adds value beyond what's already available.
The "Could ChatGPT write this?" test operationalizes that guidance into something you can actually run before you hit publish, instead of finding out three months later when the page never ranks and never gets cited.
How to Run the Test
Step 1: Write the generic prompt version in your head. Take your article's core topic and phrase it as a plain prompt, "write a blog post about X for Y audience." Now honestly imagine what a competent AI system would produce from that prompt alone, with no additional input from you. This is your baseline: the content that already exists, implicitly, inside every large language model trained on the public web.
Step 2: Compare your actual draft to that baseline. Read your draft next to the imagined baseline. Where do they diverge? Most drafts diverge only in phrasing and structure, same claims, same examples, same conclusions, different sentences. That's the failure mode this test is designed to catch.
Step 3: Identify what's missing using the four-ingredient checklist below.
Step 4: Add the missing ingredient, or kill the piece. If you can't add a real ingredient within a reasonable amount of effort, don't publish a placeholder version. A generic page doesn't just fail to help you, Search Engine Land has reported that at scale, low-value, interchangeable content on a domain can drag down how AI crawlers and search systems evaluate the rest of the site, not just the individual page.
The Four Originality Ingredients
1. A Real Opinion
Generic AI output is built to be balanced, hedged, and inoffensive by default. It presents "some experts say X, others say Y" instead of taking a position. A real opinion, "here's what I think, and here's why I think the common advice is wrong", is immediately distinguishable from generated consensus content, and it's exactly the kind of statement that gets quoted because it's attributable to a specific point of view.
2. Original Data
This is the strongest ingredient because it's the hardest to fake and the easiest for an AI system to recognize as genuinely new information. Original data can be a client result, a test you ran, a survey of your own audience, or even a simple before/after number from your own work. Semrush and Ahrefs have both published research showing that pages citing specific, sourced numbers consistently outperform pages making the same claims in vague terms, both for traditional rankings and for AI citation likelihood.
3. A Specific Experience
"I ran this exact strategy for a client in this exact situation and here's what happened" cannot be generated by a model that has never done the thing. Specificity is the tell, dates, numbers, named contexts (even anonymized), and the messy details that only show up when something actually happened. Generic content stays abstract because abstraction is safe; specific experience is what makes content unreproducible.
4. A Contrarian but Defensible Angle
Not contrarian for its own sake, contrarian because you've actually seen the common advice fail, or seen a better approach work. This is different from "hot takes." It requires you to actually disagree with the consensus and be able to explain why, with evidence, not just assert the opposite for engagement.
The Scoring Rubric
Score your draft 0-4, one point for each ingredient genuinely present (not just mentioned in passing, genuinely load-bearing to the piece's core argument):
| Score | Verdict | Action |
|---|---|---|
| 0 | Pure generic content | Kill or fully rewrite. Do not publish. |
| 1 | One weak signal of originality | Rewrite to strengthen that one ingredient before publishing. |
| 2 | Reasonably differentiated | Publishable, but look for a quick win to push to 3. |
| 3 | Strong, clearly original | Publish. This is citation-worthy content. |
| 4 | Highly original across multiple dimensions | Publish and consider this a flagship/pillar candidate. |
Be honest scoring your own work. The most common failure in running this test is grading generously because you know the effort that went into the writing, formatting, and research, none of which counts toward this score. This test measures whether the substance is reproducible by a machine that has never done any of that work, not whether the writing is polished.
What to Do With a Score of 0 or 1
Don't publish it as-is, and don't just add more words to it, padding a generic piece with more generic content doesn't change its score. Instead:
- Go find the original data point you're not currently using. Check past client work, past campaigns, past test results.
- Interview yourself (or a colleague) for five minutes on record and pull direct quotes with an actual opinion into the piece.
- If there's genuinely no original angle available on this topic right now, shelve the piece until there is one. A content calendar gap is a much smaller cost than a domain full of interchangeable pages.
This is a harder discipline than it sounds, because it means saying no to publishing volume in favor of publishing density. But that trade is exactly what's rewarded in an environment where AI systems are actively filtering for the content they don't already know.
FAQ
Does this test mean I can't use AI tools to help write? No, this test is about substance, not production method. AI-assisted drafting is fine as long as the finished piece contains a genuinely original ingredient a generic prompt couldn't have produced on its own.
What if my topic genuinely has no room for a contrarian take? Not every piece needs a contrarian angle, you only need one of the four ingredients, not all four. A strong original data point or a specific real experience is often more valuable than a forced contrarian position anyway.
How often should I run this audit? Run it on every piece before publishing, and periodically re-run it on your existing back catalog, older comprehensive content is often the most likely to score 0 or 1 because it was written before this problem existed at scale.
Isn't this just "add E-E-A-T signals," rebranded? It's a practical, executable version of that idea. E-E-A-T is a framework; this is a five-minute test you can actually apply to a specific draft before it goes live.
Can a whole site fail this test, not just individual posts? Yes, and it's worth checking. If most of your published content scores 0-1, that's a sign your content operation is optimized for volume over originality, worth auditing at the site level, not just post by post.
If you want help building an originality-first content system instead of a volume-first one, that's the kind of audit and strategy work I do at younusfardeen.com.