Reading a ghostwriter's samples when the portfolio cannot be verified

A ghostwriting portfolio is unverifiable by definition

Every page written about hiring a ghostwriter tells you to check the portfolio, and none of them acknowledges the obvious problem with the instruction. Ghostwriting is unattributed work. The whole product is that nobody knows who wrote it. A portfolio of ghostwritten posts is therefore a set of claims about authorship rather than evidence of it, and reading those posts more carefully does not convert one into the other.

The practical consequences are worse than they sound. A screenshot costs nothing to produce and nothing to borrow. A logo wall proves that money changed hands with a company, not that the person in the meeting drafted anything. An agency portfolio may contain work from writers who left two years ago. None of that requires anybody to lie, which is precisely why it is so easy.

The stakes justify the effort. Statista reports that 53% of B2B professionals rank LinkedIn as the most important social platform, which means the feed you are hiring somebody to write is, for a lot of your buyers, the main body of evidence about how you think.

So stop collecting samples and start grading evidence. The five grades below are ordered by what each one can actually prove, and the ordering is the useful part, because most hiring decisions are made on the bottom two rungs.

The Attribution Ladder: five grades of evidence

Each rung proves strictly more than the one below it. The rule that follows from the ladder is simple: never decide on rungs one and two alone, and treat a single piece of rung four or rung five evidence as worth more than a portfolio of rung one.

Rung four is the one almost nobody asks for and the one every genuinely working writer can supply within a day. That asymmetry makes it the single most efficient question in a hiring call, because the request itself is diagnostic before the material even arrives.

What an NDA legitimately hides, and what it gets used to hide

Confidentiality is a real constraint in this field and it is also the most convenient available excuse, so you need a way to tell the two apart. The distinction is that a genuine confidentiality obligation restricts identification, while it rarely restricts demonstration.

Separating a genuine confidentiality constraint from an evasion.

Redaction is the test that separates the categories, because it costs a working writer about ten minutes and offers an evasive one nothing to hide behind. Ask for one transcript excerpt with names removed and figures altered, beside the post it produced. The confidentiality objection survives that request only if the material would identify the client by its content alone, which happens, and a writer who explains exactly why is giving you a real answer.

Your own agreement will contain the mirror image of this, deciding what the writer may later say about working with you. That is worth settling deliberately rather than by default, and it sits alongside the contract clauses worth reading closely before anybody signs anything.

The live writing test, and how to run one in twenty five minutes

The live writing test is the only evidence that cannot be borrowed, and it fits inside a normal hiring call. It is not a trick and it should not be sprung on anybody. Tell the candidate it is coming, offer to pay for the time, and run it the same way with every finalist so the results are comparable.

Run this with two finalists rather than five. Forty minutes of rung five evidence resolves more than an afternoon of reading, and it has the additional benefit of showing you what working with each person actually feels like, which no portfolio can carry.

The three-voice test, scored rather than felt

Ask for three samples written for three different clients, and check whether they sound like three different people. This is the one genuinely good idea circulating on this subject, and it is always presented as an impression. Scored across four axes it becomes a usable instrument, because an impression cannot be compared between candidates and a score can.

Score each axis as same or different across the three samples. The threshold is at the bottom.

The threshold is two. Fewer than two axes varying meaningfully across three different clients means you are looking at a house style rather than three voices. Three or four axes varying means the writer listens and adapts, which is the skill you are paying for. Two exactly is worth a direct question about how they decide when to adapt and when to apply what works.

Where to spend your reading minutes

Reallocate the time rather than adding to it. The arithmetic below uses assumptions you should replace with your own, and its purpose is to show that the standard hiring process spends its attention on the weakest available evidence.

Assume you shortlist five writers and each sends three samples, which is fifteen posts. A careful read is about five minutes, because you read once for the argument and once for the sentences. Fifteen samples is therefore seventy five minutes of your attention, and all of it lands on rung one of the ladder, where nothing can be proven.

Now compare the alternative. Two finalists, twenty minutes of live writing each, is forty minutes and produces rung five evidence. Add one reference call at fifteen minutes and one redacted input and output pair read at ten minutes, and you have spent sixty five minutes to reach rungs three, four and five. That is ten minutes less than the reading, on evidence that can actually settle the question.

The same hour of attention, allocated two ways.

Read three samples, not fifteen. Three is enough to run the voice test and to decide whether the prose is worth anybody's time. Everything beyond three is reading for reassurance rather than for information, and the decision about whether this whole exercise is worth funding belongs in a different conversation about whether ghostwriting is worth the money.

What to read for once you trust the authorship

Once authorship is settled, read for four things and ignore everything else. Engagement numbers on a sample tell you about that client's audience rather than about the writing, so they are the least informative element on the page despite being the most visible.

1. Does the first line make a promise the post keeps. A hook that oversells is a small dishonesty repeated weekly, and it teaches an audience to discount everything the account says.
2. What happens in the third paragraph. Weak writers restate the hook there because they have run out of thinking. Strong writers introduce the complication that makes the argument worth reading to the end.
3. What the post does with a number. Decoration means the number sits there unexplained. Craft means the number is placed where it changes the reader's mind, and the sentence after it says what it changes.
4. Whether the post survives being summarised. If a one line summary is as useful as the post, the post is a summary, and you are about to pay a monthly fee for summaries of your own opinions.

Read one sample aloud. Ghostwritten prose that fails aloud almost always fails on the page too, because the tells are the same: clauses that cannot be spoken in one breath, transitions nobody says, and a rhythm that belongs to no particular human being.

The reference call question that gets a real answer

Never ask whether the client was happy, because the answer is yes and it carries no information. Ask about scope, time and failure instead, since those three produce specifics that a rehearsed reference cannot smooth over.

Then ask for something almost nobody asks for: one client whose engagement has ended, on good terms. A writer who can produce a former client is showing you the exit as well as the entrance, and the exit is where you learn whether the assets were handed over, whether the notice period was honoured and whether the relationship survived the invoice stopping.

If no former client can be produced after several years of practice, that is worth one direct question rather than a conclusion. There are honest answers, including a very small roster and a run of engagements that simply have not ended. There is also a less comfortable answer, and asking is how you find out which one you are hearing.

What a sample cannot tell you, however carefully you read it

A sample is a photograph of a finished thing on a good day, and four of the qualities that decide whether an engagement works cannot appear in one. Knowing which four keeps you from over-reading the evidence you have.

• Reliability under a bad week, which only shows up when something goes wrong on either side and there is a deadline anyway.
• How they take an edit. A writer who defends every sentence is expensive in a way no sample reveals, and a writer who accepts every change is not editing you at all.
• Whether the interest survives month five, when the obvious arguments have been made and the work becomes finding new ones.
• Whether they can write about your particular subject, which is a research question rather than a prose question, and research capacity is invisible in finished work.

All four are only observable over time, which is the argument for a short paid engagement with a stated end rather than a longer sample process. The design of that engagement decides whether it answers anything, so it is worth setting up a trial period with a defined kill point rather than a discounted first month that quietly becomes month one of a year.