What a hit actually means, and the four false positives
A match is a question, not a verdict. Four things produce hits that mean nothing, and it's worth ruling them out before you send an email you can't take back.
- Scraper sites. Your own article, republished wholesale on a domain you've never heard of, usually with the internal links intact. That's theft from you, not by your agency — and it's common enough that it should be the first thing you check.
- Your own staging or dev subdomain.
-site:yourdomain.comdoesn't excludestaging.yourdomain.com. If a hit is your own copy on a subdomain, you have an indexing problem rather than a plagiarism problem, and it's costing you. - Genuinely shared source material. A manufacturer's product specification, a statutory definition, a GST rate, a quote you both took from the same Google documentation. Identical because reality is identical.
- Syndicated PR. If your press release went out on a wire, it exists verbatim on forty sites by design.
Reuse that's legitimate, and reuse that isn't
There is a version of this check that turns into a witch hunt, and it makes your agency worse rather than better. Every competent content team reuses something. A research method, a brief template, a way of laying out a comparison — those are the accumulated craft you're paying for, and demanding that each article be invented from nothing produces slower, sloppier work.
The line sits at the words and the specifics. Google's spam policies name scaled content abuse — many pages generated mainly to manipulate rankings rather than to help anyone — as a violation, and reselling one article to six clients with the nouns swapped sits squarely inside that definition for at least five of them.
| Reused element | Fine, and why | Not fine, and why |
|---|---|---|
| Research process | The SERP check, the interview questions, the source-quality bar. Craft, refined across accounts. | Nothing here is a problem. If they don't have one, that's the finding. |
| Page structure | A comparison page that always carries a table and a verdict is a house style, not laziness. | The same nine H2s in the same order across four clients in the same industry. |
| Examples and numbers | A public benchmark or a documented rule, cited, is the same everywhere. | A worked example about a Bangalore SaaS company in an article you bought for a Coimbatore textile mill. |
| Sentences | Boilerplate you supplied — your legal line, your product description. | Any body paragraph that appears on another company's website. There's no benign version of this. |
| The brief | A repeatable brief format speeds everything up and makes review easier. | A brief with no customer input, no SME name and no query research behind it. |
Was it written for your buyer? Three questions any reader can apply
Originality is the low bar. An article can be 100% unique and still be worthless, because uniqueness is trivially easy to manufacture and usefulness isn't. Google's own content self-assessment asks whether a page offers original information, reporting, research or analysis — not whether the string of words is new.
These three questions need no tool and no SEO knowledge. Anyone in your company who talks to customers can run them in five minutes per article.
- The logo swap. Cover your brand name and read it. Could your closest competitor publish this, unedited, with their logo on it? If yes, it isn't your content — it's category content, and it will be outranked by whoever has more links.
- The one thing only you know. Find the sentence containing a fact that could only have come from your business: a real price, a lead time, a constraint, an objection your sales team hears every week, a reason you turn certain customers away. If there isn't one, no research happened. This is the single most reliable tell in the whole exercise.
- The buyer's own words. Does it use the vocabulary your customers actually use, or the vocabulary of a keyword tool? For an Indian audience that includes the boring specifics: rupees rather than dollars, GST rather than sales tax, WhatsApp as a real channel, city names your customers would recognise.
The monthly check, in about twenty minutes
Do this on the day the invoice arrives, on a sample rather than everything. The point is deterrence more than detection — an agency that knows two random articles get checked every month behaves differently from one that doesn't.
- Pick two articles at random from the month's delivery. Random matters; if you always check the first one, that's the one that gets written properly.
- Run the sentence search and the skeleton search on both. Four searches, four minutes.
- Apply the logo swap and hunt for the one thing only you know. Ten minutes.
- Check the internal links. An article written by someone who read your site links to your service pages using anchor text that matches what you sell. An article written by someone who didn't links nowhere, or links to Wikipedia.
- Check it's indexed. Search a phrase from it in quotes with
site:yourdomain.com. Published and unindexed after three weeks is its own conversation, and it's usually about the site rather than the writing.
You found something. What to do short of terminating
Firing an agency is expensive, slow and frequently the wrong response to a fixable process failure. Unless you found verbatim reuse across clients — which is a trust problem, not a process problem — work the ladder before you work the notice period.
The whole exercise gets easier if the contract said something about it upfront. Originality warranties, a named writer, and the right to a source list are three clauses that cost nothing to add and change the incentives entirely.
- Ask for the brief and the source list for that specific article. Not a template — the actual brief. A team doing the work has both in a folder. A team resurfacing an old draft has neither, and the delay in answering tells you more than the answer does.
- Ask who wrote it, by name, and ask to speak to them. Reasonable request, ten-minute call. It also surfaces the outsourcing question without you having to accuse anyone of anything.
- Insist on a subject-matter interview per article. Thirty minutes of your sales lead's time per piece is the cheapest quality fix available, and it makes the logo-swap test almost impossible to fail.
- Cut the volume and raise the depth. Six shallow articles a month reliably lose to two researched ones. If the retainer is fixed, renegotiate what it buys rather than what it costs — and check the numbers against what content actually costs to produce in India so the trade is honest on both sides.
- Put a rewrite obligation in writing for anything that fails the exact-phrase test, at their cost, within ten working days. Most agencies will agree to this because most agencies aren't doing it.