Did AI Write It LogoDid Write It

You Paid for Content. Did You Get AI Output?

If you commission writing from an agency, a freelancer, or a contractor, the risk is not the invoice. It is publishing thin generated copy under your own brand. Here is how to check what you received.

AI Detection4 min readBy Curtis Nye
Content MarketingFreelancePlagiarismSEO

A writer who used to spend six hours on a 1,500-word article can now produce something structurally similar in twenty minutes. The economics changed fast, and quietly. Most writers haven't lowered their rates to match, and there's no reason they'd volunteer that they could.

None of that is automatically a problem. A writer who runs a model to sketch an outline, then does the real thinking and the actual editing themselves, is still doing the job you hired them for. What you actually want to catch is different. It's a deliverable that got generated, then dressed up and billed as if someone wrote it fresh.

Why this is your problem, not the contractor's

The instinct is to frame this as a money question, whether you got what you paid for. That's the smaller issue.

Commissioned content publishes under your brand, so whatever's wrong with the piece becomes a cost you carry long after the contractor has cashed the check and moved to the next client. Google's own guidance already discounts unhelpful, mass-produced content no matter who or what wrote it, so thin generated copy is a bad investment even when no human reader ever notices, since you're paying for an asset that just won't rank. The sharpest risk is the one that turns legal instead of stylistic: if a deliverable lifts real passages from something nobody licensed to you, that exposure attaches to the publisher. That's you, not the contractor who already moved on.

What to actually check

Three checks cover the realistic failure modes here, and all three run on the same pasted text.

  • AI detection, for whether the prose reads machine-generated. Read the sentence-level report instead of the single overall number. Light AI assistance and a wholly generated draft look genuinely different once you're looking at individual passages instead of one score.
  • Plagiarism, for overlap with sources already published elsewhere. This one returns the matched sources and highlights the exact overlapping text, which is the part you'll actually need if you ever have to show someone.
  • Fact-checking, for claims that nothing backs up. Each one comes back labeled with its own verdict and source links, rather than smoothed into one aggregate number.

It's worth running all three on content you already published, not just new drafts. The plagiarism check especially. Anything already indexed by Google is the highest-exposure content you own. It's also usually the easiest to check, since it's already sitting on your own server.

Put it in the contract first

A report is far more useful as evidence against a written clause than as an argument in the absence of one. If originality matters to you, say so before the work starts. Say specifically what you mean, too. Rule out wholly AI-generated drafts, require disclosure whenever assistance was used, and ban invented sources or citations outright.

Skip that step and a flagged deliverable puts you in an argument about whether a probabilistic score proves anything. You'll lose that argument, because it genuinely doesn't. Do that first, and the same report becomes proof of a specific breach, not a vague number.

Reading the result fairly

A high score is a signal, not proof of anything on its own. Detection misfires in a few well-known ways. Short passages trigger easily. Heavily templated formats do too. So does writing by non-native English speakers, whose careful, formal phrasing can read to a detector almost exactly like a model's. A contractor who happens to write clean, plain, efficient prose can get flagged honestly, and treating that flag as a confession will cost you a genuinely good writer.

The better move is to send the report itself rather than a number. Every saved report has a read-only share link, so you can point to the specific flagged passages and ask about those directly instead of paraphrasing a score at someone. A writer who actually did the work can usually talk you through it sentence by sentence. That conversation resolves most of these faster than any score will on its own.

To check something you've been sent, the content verification page runs a free scan with no account needed. If you're the one on the other side of this, and you need to show a client that your own work is actually yours, proving you wrote it covers that instead.

Scan your draft before you publish.

See what gets flagged, fix what matters, and rescan to compare versions.

Start free