How to Onboard a LinkedIn Ghostwriter: The First Thirty Days

On this page

A ghostwriting portfolio proves a file exists, not who wrote it
The Personeur Attribution Test: five checks that beat a portfolio
The three voices test, run properly
Days one to ten: the deliverable is the voice file, not the first post
Days eleven to twenty: keep an edit ledger and read what your edits mean
Days twenty one to thirty: publish, and measure what a month can actually measure
The day thirty decision, scored
What to write into the trial agreement so day thirty is clean

The short answer

A ghostwriting portfolio is unverifiable by construction. The work is unattributed, samples are often collaborative, and nothing distinguishes a piece the writer produced from a piece the client largely wrote. Replace portfolio review with evidence you can check: the raw input beside the output, a paid live exercise using your own material, three samples tested for three different voices, and a structured thirty day trial with an edit ledger and a scored exit.

A ghostwriting portfolio proves a file exists, not who wrote it

Every hiring guide tells you to check the portfolio, and none of them acknowledges the problem sitting in plain view. Ghostwriting is unattributed by definition. The finished piece carries someone else's name, so a sample can only ever be a claim of authorship rather than evidence of it.

Three specific things go wrong with samples in this craft. Work gets borrowed from a former employer or a subcontractor and shown as personal work. A piece the client substantially rewrote gets presented as the writer's output. A strong sample turns out to have been produced under a senior editor who is no longer involved. None of these require anybody to be dishonest, which is what makes portfolio review so unreliable.

53%

53% of B2B professionals rank LinkedIn the most important social platform for their work, which makes an executive account a business asset rather than a personal page. Handing it to someone on the strength of unverifiable samples is the part of this process nobody defends out loud.

Statista, 2026

The answer is not to distrust writers. It is to stop asking a portfolio to do a job it cannot do, and to move the verification into the first thirty days, where you can watch the work being made from material you supplied yourself.

The Personeur Attribution Test: five checks that beat a portfolio

Each check below produces evidence tied to a person rather than to a file. Run at least three of the five before you sign anything longer than a trial.

The Personeur Attribution Test

Five checks that survive the fact that the byline is not theirs.

The input beside the outputAsk for one published piece together with the raw material it came from: the transcript, the voice note, the scribbled brief. Names and numbers can be redacted. A writer with a real process has these files as a by product of the work and can produce one in a minute. A refusal of the format itself, rather than of the details, is the answer.

The paid live exerciseSend twenty minutes of recorded conversation about something you actually think, then commission one short piece from it, and pay for it. Watch whether your own phrases survive into the draft or whether the material has been poured into a template. This is the single most informative hour in the whole hiring process.

The three voices testAsk for three samples written for three different clients and check whether they sound like three different people. This test is the best idea on the subject and it is usually mentioned in one line and never explained, so the next section sets out exactly what to look for.

The confirming clientAsk for a client who will confirm the working relationship without confirming which posts were ghostwritten. That separation respects the confidentiality that genuinely binds this craft while still giving you a human being to speak to. Discretion explains why the posts stay private and never why nobody will take your call.

The version trailAsk to see a first draft beside the version that published, on any piece, in any redacted form. Real drafting leaves a trail of cuts, reordered arguments and abandoned openings. A writer who only ever has final files has either lost the habit of revising or was not the one revising.

The paid exercise deserves the emphasis. Paying for it changes what you are entitled to expect and removes the awkwardness of asking for free work, and a writer who declines a paid test on your own material is telling you something the portfolio never would.

The three voices test, run properly

Three samples for three clients either sound like three people or they sound like one writer with a house style and three sets of nouns swapped in. Reading them casually will not tell you which. Read them against these five markers instead, and score each marker as differentiated or shared.

The opening move. Note how each piece enters: a question, a flat claim, a scene, a number. A writer with one voice has one entry and uses it everywhere, so three identical entries is the strongest single signal available.
Sentence rhythm. Count the words in the first five sentences of each sample. A shared cadence pattern across three clients is the writer's rhythm rather than the clients', because real people vary enormously in how long their sentences run.
The tic inventory. List the recurring devices: the one line paragraph before the point, the colon in the opening line, the numbered list arriving in the same position, the closing question. Tics migrate between clients when one person is writing all three.
Vocabulary sourcing. Check whether each sample contains language only that industry uses, deployed correctly, or whether the sector appears purely as nouns dropped into general business prose. The second pattern means no interview happened.
What each one disagrees with. Three clients should be contrarian about three different things. When all three push back on the same category of received wisdom, you are reading the writer's opinions in three costumes.

Score it simply. Three or more markers clearly differentiated is a pass. Two is a conversation worth having, in which you ask directly how much of the voice work is deliberate. Fewer than two means you would be buying that writer's voice under your own name, which is a legitimate thing to buy and should at least be a decision you made knowingly.

A fair caveat

Some writers genuinely specialise in one sector and one seniority, and their samples will share vocabulary for honest reasons. Read the markers as a whole rather than failing anybody on vocabulary alone. Rhythm and the opening move are the two that are hardest to fake and hardest to explain away.

Days one to ten: the deliverable is the voice file, not the first post

Nothing needs to publish in the first ten days. What needs to exist by day ten is a written voice file you could hand to a different writer, because that file is the asset the engagement is really building.

Day one, hand over raw material rather than a brief

Send recordings and documents you did not write for this purpose: two customer calls, an internal memo, a talk you gave, the message you sent when something went wrong. Material written for the writer describes the person you think you are. Material written for other reasons contains the person you are.

Days two and three, the recorded interview

Sixty to ninety minutes, recorded, with the writer asking and you answering. If the writer sends a questionnaire instead of booking time, the engagement has already defaulted to assembling your public material. Ask for the transcript afterwards and keep your own copy.

Days four to six, the voice file gets written

It should contain your positions in your own words, the examples you return to, the phrases you actually use, a list of words you would never say, and the two or three things you are known to argue with. Read it and correct it. This is the hour of your time with the highest return in the whole engagement.

Days seven to ten, three drafts against the file

Three pieces, unpublished, written from the interview rather than from your old posts. Read them against the voice file rather than against your taste, because taste on a Tuesday is not a standard anybody can write to twice.

Day ten, agree the approval mechanics

One named approver, a stated response window, and a rule for what happens to a draft that misses its window. Approval design decides your publish rate far more than drafting talent does, and leaving it undefined is how a good trial produces four published posts.

Format decisions belong in this window too, since the writer's strengths differ by format and the trial should test the formats you will actually publish. The tradeoffs are set out in choosing between text, image and document posts, and picking two formats for the trial is better than sampling all of them badly.

What to take away

A ghostwriting portfolio proves a file exists, not who wrote it, because the craft is unattributed and clients routinely rewrite the pieces they are shown as samples.
The checks that survive scrutiny are the raw input beside a published piece, a paid live exercise on your own material, the three voices test, a client who will confirm the relationship without confirming the posts, and a first draft shown beside the version that published.
Keeping a ledger of what you changed each week tells you where the failure sits, because factual edits mean a broken intake and word level edits mean a thin voice file.
Thirty days is too short to judge reach and exactly long enough to judge fit, so measure the edit pattern rather than the impressions.
The ethical line is not that someone else typed it. It is publishing something you cannot remember approving and would not defend in a meeting.

Days eleven to twenty: keep an edit ledger and read what your edits mean

Every edit you make is a diagnostic signal about which part of the system is failing, and almost nobody records them. Keep a one line note for each draft naming what you changed. By day twenty the pattern is unmistakable, and the pattern is what you act on.

What you changed	What it actually means	Where the fix belongs
Facts, names, numbers, the order events happened in	The intake is broken. The writer is working from inference rather than from your material	The interview and the source pack, never the drafting. Send raw documents and stop briefing verbally
Individual words and phrasings, with the sentences left standing	The voice file is thin on vocabulary. This is the cheapest failure to fix and the most common in week two	Add a preferred phrase list and a banned word list to the voice file, both drawn from your own transcripts
The structure, moving the point earlier or later	A craft gap on format rather than a voice gap	The writer's own editing standard. Show them two of your posts that worked and name what the opening does
The argument itself, what the piece claims	The angle was approved loosely or never approved at all	The idea queue. Approve one sentence angles before drafting begins, not topics
Nothing, but you feel uneasy publishing it	The draft is defensible and not yours. This is the failure that survives thirty days undetected	Read the line aloud. If you would not say it in a meeting, it is not your voice whatever the file says
You rewrote it completely and published your own version	Count these. Two in a month is calibration and four is a mismatch of judgement	A structural conversation at day thirty, not another revision round

The edit ledger. What you changed, what it means, and where the fix belongs.

The direction of travel matters more than any single week. A healthy trial moves down that table: factual corrections in week one, word level corrections in week two, structural notes in week three, and very little by week four. A trial still stuck on factual corrections in week four has an intake problem that drafting talent cannot reach.

Days twenty one to thirty: publish, and measure what a month can actually measure

Publish in this window and resist the urge to judge the engagement on reach. A month of posting cannot separate a writer's quality from timing, topic and the ordinary variance of a feed, and no published benchmark repairs that. What a month can measure is fit, and fit has three observable signals.

The edit trend from your ledger, which is the cleanest signal you have and the only one entirely inside your control.
Who replies rather than how many. Ten replies from people who look nothing like your buyer is a worse month than two replies from the exact roles you sell to, and the difference will not appear in an impressions figure.
Whether you would have published it yourself. Ask the question about each piece before you look at any numbers, and record the answer. A month of yes means the voice work landed, whatever the feed did that fortnight.

On judging reach in month one

Reach in a single month is dominated by factors the writer does not control: what else your audience saw that day, how your account has behaved historically, and the ordinary swing between one post and the next. Anyone who guarantees a reach outcome in a thirty day trial is selling certainty that does not exist. Judge the fit in month one and the results across a quarter.

One habit is worth starting in this window and keeping permanently. Save the drafts you rejected, with the reason attached. Six months later that folder is the most accurate description of your standard that anybody has ever written down, and it is the fastest way to bring a second writer up to speed.

The day thirty decision, scored

Make the decision on the ledger and the attribution evidence rather than on how much you liked working together, and make it on the day rather than letting the trial drift into a contract by inertia.

Signal at day thirty	Reading	Decision
Your edits moved from facts, to words, to almost nothing	The system is learning in the right order	Continue, and lengthen the approval window now that trust exists
Still correcting facts in week four	The intake never really happened, and drafting talent cannot reach this	One more cycle with raw material supplied directly, then stop if it repeats
Edits are all at word level and stable	Good fit with a thin voice file	Continue, and spend one hour adding vocabulary and examples to the file
You rewrote three or more pieces yourself	A mismatch of judgement rather than of effort	Stop. This does not improve with time and both sides usually know before you say it
You published everything and cannot recall what any post argued	You have outsourced the thinking rather than the writing, which is the actual line	Pause publishing and reset the approval process before anything else continues
The attribution checks were never completed	You are thirty days in and still relying on a portfolio	Complete them now. A writer inside a live engagement has less reason to decline, not more

Read the signal from the trial, then take the decision beside it.

The fifth row is the one worth sitting with. It is not a quality failure and it will not appear in any metric. It is the point at which a convenient arrangement has quietly become a misrepresentation, and it is treated at length in whether ghostwriting is dishonest.

What to write into the trial agreement so day thirty is clean

A trial that ends without argument is one that stated its terms at the start. Six lines cover almost everything that goes wrong, and four of them are about ownership and accountability rather than money.

The six lines a thirty day trial needs

A fixed end date and a fixed fee, with no automatic continuation. A trial that renews by silence is not a trial.
Ownership of everything produced, assigned on payment, including the transcripts, the voice file and the rejected drafts. The voice file is the asset that makes your next writer cheaper.
One named person accountable for every factual claim in a published post, and that person is you. A writer can research, check and challenge, and the account is yours, so the final responsibility cannot be delegated with the typing.
A written rule against publishing metrics or claims without a source you have seen. This is the clause that prevents a plausible number appearing in your name and staying there.
An approval trail: who approved what, and when, kept in one place. Regulated businesses need it as a matter of record and everybody else needs it the first time a post causes a problem.
Your agreed answer to the question of whether you write your own posts, decided before anyone asks rather than in the moment. Deciding it in advance is what makes the answer calm.

That last line reads like a small thing and it is the one people regret leaving open. A position agreed on day one is a policy, and the same words invented on the spot sound like an admission. The contractual and legal side of the same question, including who holds copyright by default and what an assignment clause needs to say, is covered in where the law actually sits on ghostwriting.

The portfolio tells you what a finished piece looks like in that shop. The first thirty days tells you whether the thing they build is yours, and only one of those two is checkable.The house position

Questions people ask next

Is it unreasonable to ask a ghostwriter to prove they wrote their samples?

It is standard practice in every other unattributed craft, and the request is normally welcomed by people with a real process. Ask for one redacted input file beside one published piece rather than for proof across the whole portfolio. That single artefact is enough, and demanding more turns diligence into suspicion.

Should a thirty day ghostwriting trial be paid?

Yes, in full and at the normal rate. An unpaid trial attracts writers with spare capacity rather than writers in demand, and it gives you no standing to ask for revisions or to expect a deadline to hold. Pay for the trial and treat it as the first month of work, which is exactly what it is.

How many posts should a thirty day trial produce?

Between four and eight published pieces is enough, and the number matters less than the mix. Include at least one piece you expect to be difficult, meaning one that takes a position somebody could argue with. Easy posts tell you nothing about voice, because generic writing passes as anybody's.

What if the writer refuses to share the raw input behind a sample?

Distinguish between refusing the details and refusing the format. Redacting a client name, a number or a whole paragraph is legitimate and expected. Declining to show any input file at all, in any redacted form, from any engagement, means either the process does not produce them or the sample was not theirs.

Can I run the three voices test on an agency rather than a freelancer?

Yes, and it means something different there. Three samples from three different agency writers should obviously sound different, so the test becomes a check on whether the agency imposes a house style on every client. Ask which writer produced each sample, then ask to see three from the one person who would be assigned to you.

What happens to the voice file if I change writers later?

It should transfer with you, which is why ownership of it belongs in the agreement rather than in an assumption. A written voice file, a set of transcripts and a folder of rejected drafts lets a new writer start at roughly month three instead of month one. Without them, every switch pays the full onboarding cost again.

Sources