You Sound Just Like the Company You Are Trying to Beat

Take your last ten posts. Cut off the byline, remove the company name, strip any first person detail specific enough to identify you. Now hand that stack to someone in your industry and ask them to guess who wrote it.

Most people fail this test with everything they have published in the last year.

The swap

Here is a faster version if you want the answer before lunch.

Take your single best performing post from the last quarter. The one you were proud of. Now imagine your closest competitor published it, word for word, under their logo, tomorrow morning.

Would anything look wrong?

If the answer is no, you did not publish thought leadership. You published a Stock Opinion. Clean, polished, and available in identical form from every other vendor in your category. Nobody licensed it to you, which somehow makes it worse.

The name matters because it points at the right fix. A stock photo is not bad photography. The lighting is good. The problem is that it belongs to nobody, so it says nothing about the person using it. That is your content when it could carry any logo in your market.

This started before the machines

The easy explanation is AI, and the numbers make it tempting. Pangram Labs published research this year, covered by The Register, finding 41 percent of LinkedIn posts over 250 words are machine written. Not assisted. Written. Highest of any platform they measure, against a 25 percent average elsewhere. Originality.ai reached a similar range in its own study using its own threshold.

But the machines did not invent the sameness. They learned it from us, and the trail goes back nine years.

In late 2017 a growth strategist named Josh Fechter worked out that LinkedIn showed only the first line of a post before a see more link, and that clicking that link registered as a quality signal. So he wrote in single sentences with a hook on top, kept links out of the body, and built every post around a personal story with a business lesson attached. BuzzFeed reporters named the style broetry and quoted him calling the method spoonfeeding people what they want. By his own account it produced 200 million views in six months, until LinkedIn adjusted the algorithm and his engagement collapsed from around 5,000 a post to 300.

The exploit died in 2018. The format never did.

Nine years later that cadence is the house style of an entire platform, and most people writing it have never heard the man’s name. Then the loop closed. Writing tools trained on nine years of that content now produce it as a default. The machines learned the shape from people who learned it from a click hack, and somewhere in there it stopped being anyone’s voice.

So the sameness is not an AI problem. AI is the accelerant on a fire that has been burning since 2017.

The incentive nobody names

There is one more finding that explains why this keeps paying, and it is uncomfortable.

Originality.ai measured engagement alongside detection. Posts flagged as likely machine written pulled 45 percent less engagement than human ones on average. One category broke the pattern. In leadership and inspiration content, the machine written posts beat the humans by 75 percent.

Read that twice. The platform discounts synthetic writing everywhere except the one genre that runs on sincerity.

A rational person takes an obvious lesson from that, and the feed shows they have. The manufactured humility, the airport realization, the story about the candidate who cried, all of it keeps arriving because it outperforms the real thing. That is not a moral failure spreading through your industry. It is a price signal, and people follow price signals whether or not they admit it.

Which means the sameness is rewarded, up to the point where it stops working, and nobody can tell you where that point is.

What it costs you

The spend attached to Stock Opinions is bigger than most teams admit.

The agency retainer producing weekly posts. The ghostwriter on the executive’s byline. The content calendar with twelve slots a month that must be filled whether or not anyone has anything to say. The tool subscriptions that turn one bullet into four paragraphs. And the executive hours spent reviewing drafts that could have carried a competitor’s logo.

Then the cost that never gets counted. Every generic post trains your audience that your name predicts nothing. Once someone learns your byline carries no information, the good post you write six months from now does not get read, because they stopped checking. You did not just waste the budget on the weak pieces. You spent it teaching people to skip you.

There is a compounding version too. I argued in the case study piece that buyers now assemble their verdict from sources you do not control. Systems doing that assembly weigh distinctiveness. Content that reads like every other vendor gives them nothing to attribute to you, which is one reason a company can publish every week and still go missing from the answer, a gap I covered in the AI search piece.

Why the swap test keeps failing

Nobody sets out to write a Stock Opinion. The process produces them.

Content gets approved by committee, and committees remove the parts that stick out, because the parts that stick out are the parts someone might object to. The first thing cut from any draft is the specific number, the named mistake, the position a customer might disagree with. What survives review is what nobody could object to, which is another way of saying what nobody could remember.

Then there is the safety of the middle. A strong claim can be wrong in public. A vague one cannot. So the incentive for the individual writing it points toward hedge, and the incentive for the company approving it points toward hedge, and the output is two parties agreeing to say very little at length. This is the same failure I described in how good strategy gets killed, showing up at the level of a sentence.

The tell is easy. Read your last post and count the claims a competitor would refuse to make. If the number is zero, you wrote wallpaper, and the review process worked as designed.

There is a second force behind it that nobody says out loud. Most B2B content gets produced on a calendar rather than on an idea. Twelve slots a month exist, and they must be filled whether or not anyone in the building has learned something worth reporting. A person with a real insight writes one thing. A person with a quota writes twelve, and eleven of them are shaped like the first, because the shape is what remains when the substance runs out.

The way out

Publish something that costs you.

That is the whole method and it is harder than it reads. A position that can lose you a deal, a number from your own operation, a mistake with your name on it, a prediction with a date attached that someone can check. Anything with a price on it cannot be swapped onto a competitor’s site, because they would have to pay the same price to say it, and they will not.

Three things that pass the swap test.

Your own data. Numbers from inside your business cannot be duplicated by anyone. A study of fifty websites, your own conversion figures, the results of a test you ran. Nobody else can publish it because nobody else has it.

A named enemy. Not a competitor, a practice. Say which common approach in your category is wrong and why, in terms specific enough that half the market would refuse to sign it. Agreement is what makes content interchangeable.

A dated prediction. Say what happens by when, publish it where it can be checked, and revisit it in public. It is the cheapest form of skin in the game and almost nobody in B2B will do it.

Then run the swap test before you publish, not after. One question, ten seconds. Could my closest competitor put their logo on this tomorrow?

If they could, they will not have to. Somebody in your category is publishing that exact post this week, and their audience is skipping it too.