Checklist · 8 min read

Three minutes between draft and damage.

Most posts that hurt your credibility were not badly written — they were unchecked. This is the pass to run on every draft before it goes out: eight groups, twenty-one checks, and the exact condition that makes each one fail.

Quick answer

What should you check before publishing a thought leadership post?

Check eight things: it came from real source material, it argues one claim, every fact is defensible, it sounds like you, the first two lines work truncated, it is scannable, it ends on the claim rather than an engagement plea, and nothing in it is invented. The whole pass takes under three minutes on a finished draft.

One claim per post — a list of five is not a position.
Cut any number you cannot source on request.
No composite customers, ever.
The first two lines must work cold and truncated.
End on the claim, not on 'Agree?'.

The checklist at a glance

f you only remember eight lines, remember these. Each maps to a group below, and each is phrased as the condition that must be true — so a failed check tells you precisely what to fix rather than sending you back to rewrite the whole post.

Check

Must be true

Source

Came from a real moment you can name.

Group 1

Claim

One arguable position you'd defend on a call.

Group 2

Evidence

Every number sourced; every story real.

Group 3

Voice

No word you wouldn't say out loud.

Group 4

Hook

First two lines work truncated and cold.

Group 5

Shape

120–250 words, scannable, no stunts.

Group 6

Ends on the claim, not on 'Agree?'.

Group 7

Honest

Nothing invented, nobody exposed.

Group 8

Chapter 01

The eight groups

Work top to bottom. The early groups are the expensive ones — a post that fails group one or two cannot be rescued by anything in groups five to eight.

1. Source material

The post traces back to something that actually happened in your week.
Fails if: You started from a topic rather than a note. Topic-first posts are why AI content reads as generic.
You can name the moment: the call, the objection, the deal, the question.
Fails if: The origin is 'something I've noticed generally' — that is opinion without evidence.

2. The claim

The post argues exactly one thing.
Fails if: It lists five tips. A list is content; a claim is a position, and only positions get remembered.
A reasonable person in your market could disagree with it.
Fails if: Nobody could object — 'consistency matters' is not a claim, it is furniture.
You would defend it on a call, unprompted.
Fails if: It is a take you borrowed because it sounded strong. The first hard comment will expose it.

3. Evidence

Every number in the post has a source you can produce on request.
Fails if: The statistic came from the model, or from a blog post you did not open.
Any customer story is real, and either permitted or anonymised honestly.
Fails if: It is a composite 'client of mine' assembled from several accounts.
Claims about results are stated with their limits.
Fails if: You imply attribution you cannot prove, like a percentage lift with no baseline.

4. Voice

Every sentence is one you would say out loud to a buyer.
Fails if: There is a word in there you have never used in conversation.
The adjectives are ones you would defend, not ones a model reached for.
Fails if: The post is 'thrilled', 'passionate' or 'game-changing' anywhere.
The rhythm matches your own writing — sentence length, openings, humour.
Fails if: You had to describe your voice to the model in adjectives instead of showing it samples.

5. The first two lines

The opening states the tension instead of teasing it.
Fails if: It only works if the reader taps 'see more' — that is a cliffhanger, not a hook.
Read cold, out of context, the first two lines still make sense.
Fails if: They depend on the paragraph below them to mean anything.

6. Structure

The post is scannable: short paragraphs, no wall of text.
Fails if: Any paragraph runs past four lines on a phone.
Length sits roughly between 120 and 250 words.
Fails if: It is past 300 words, which usually means it contains two posts.
No formatting stunts.
Fails if: One word per line, emoji bullets, or a fake countdown to a reveal.

7. The close

The final line is the claim, stated plainly.
Fails if: It ends on 'Agree?', 'Thoughts?' or a follow-for-more plea.
If there is an ask, it is one a stranger could act on for free.
Fails if: The close is 'DM me' with nothing given first.

8. Ethics and safety

Nothing in the post asserts an experience you did not have.
Fails if: The narrator is a slightly more impressive version of you.
No named person, customer or employer is identifiable without consent.
Fails if: Details would let a reader work out who it was.
You could sit opposite a buyer and be asked about any sentence without flinching.
Fails if: Any sentence would need explaining away. Cut that sentence, not the post.

Chapter 02

The order to run them in

Run groups one to three before you polish anything. They decide whether the post should exist; the remaining five only decide how well it lands. Polishing a post that fails the claim check is the most common way people waste their writing time.

The one-question version

If you have thirty seconds rather than three minutes, ask only this: could I sit opposite this buyer and be questioned on any sentence here without flinching? It catches invented numbers, borrowed opinions, composite customers and overclaimed results in a single pass.

A failed check is not a reason to skip publishing.

It is a reason to publish the shorter, truer version. A thin post that is entirely defensible builds more standing over a year than a polished one that a buyer later discovers was padded.

SelloWise runs a version of this automatically — every draft is scored against the SIGNAL framework before you publish, and the reasoning behind that scoring is in the Authority Playbook.

Chapter 03

Frequently asked questions

What should I check before publishing a thought leadership post?

Eight things: that it came from real source material, that it argues one claim, that every fact is defensible, that it sounds like you, that the first two lines work truncated, that it is scannable, that it ends on the claim rather than a plea for engagement, and that nothing in it is invented. If any check fails, fix that item rather than rewriting the post.

How long should the check take?

Under three minutes once you know the list. It is a pass over a finished draft, not a second drafting session. If a post regularly takes ten minutes to check, it is usually because the source material was thin and you are trying to repair that at the end.

What is the single most important check?

Whether you could defend every sentence to a buyer sitting opposite you. That one question catches invented statistics, borrowed opinions, composite customers and overclaimed results in a single pass — which is most of what actually damages credibility.

Does the checklist change for founders or marketers?

The checks are the same; the evidence differs. A seller's evidence is objections and deals, a founder's is decisions and numbers they own, a marketer's is experiments and results. The requirement — a specific, defensible source for the claim — does not move.

Should I check AI-assisted posts more carefully than ones I wrote myself?

Only on the facts. A model will produce a confident, plausible, wrong number in a way you would not, so verification deserves extra attention. Voice, claim and structure need the same scrutiny either way, because a generic post written by hand fails for exactly the same reasons.

Have the check run itself.

SelloWise scores every draft against the SIGNAL framework — claim, evidence, voice and hook — before it reaches your buyers.

Keep readingTen post templates Before-and-after examples

More guides in the SelloWise learning library.