Every asset a personal branding guide tells you to build is written in the first person. The headline, the bio, the about page, the elevator line. All of them are you, describing you.
The artifact that decides whether any of it worked is written in the third person, by somebody else, in a conversation you are not part of. It is about one sentence long and it usually begins “you should talk to.” That sentence is the entire mechanism. A referral is not a warm feeling somebody has about you. It is a sentence they can produce accurately, without preparation, while thinking about something else.
Nothing in the standard advice performs that translation. It produces better first-person copy and assumes the third-person sentence will follow. Often it does not, and the reason is specific enough to do something about.
The half the software got good at
First-person copy is exactly the work a language model does well. Give it your background and it returns headlines, two bio lengths, tagline options and a month of post drafts in a few minutes, at a standard most owners would not reach alone on a Sunday evening. That is real and it is worth using.
It also does nothing to the sentence that matters, and that is the trap rather than a complaint. Faster first-person copy gets you to a polished version of an untested claim sooner, and polish reads like progress. What you end up holding is a confident headline that nobody outside your own head has ever said out loud.
Asking the model whether the claim is any good does not rescue it either, and there is measurement behind that now. A Stanford-led study published in Science in March 2026 evaluated 11 major models and found they endorsed the user 49 percent more often than human responders did (Stanford Report). We went through what that habit costs an owner in AI Sycophancy: Chatbots Agree 49% More Than People. The narrow point here is that your own positioning is precisely the kind of question you have nobody safe to ask, which is why it ends up in a chat window, and a chat window is the last place that will tell you the sentence cannot be repeated.
Why “be specific” is only half an instruction
The standard fix for a vague brand is specificity. Stop writing “cybersecurity specialist” and start writing something closer to “I help midmarket SaaS companies cut breach response time by 60 percent.” Every guide arrives at some version of this, and as first-person copy it is a genuine improvement. It names a buyer. It names an outcome. It carries information.
Now hand it to the person who has to repeat it.
That person is frequently not your client. It is somebody who met you once, or who heard about you from somebody who did, and who has a few seconds in the middle of a conversation to produce your name and a reason for it. Faced with a number they did not measure, they have two options. They can drop it, which returns the sentence to vagueness. Or they can repeat it, which quietly makes them responsible for a claim they cannot defend if the other person pushes back. Most people drop it, and they are right to.
There is a rough ceiling on what anybody carries out of a conversation at all. Nelson Cowan’s review of the short-term memory evidence puts the working capacity at about three to five chunks rather than the seven that entered folklore (The Magical Number 4 in Short-Term Memory, Behavioral and Brain Sciences, 2001). That is a laboratory finding about memory rather than a rule about referrals, and it should not be stretched into one. It is still the only honest reason anyone has offered for stopping at three clauses instead of five.
So the target is not the most informative sentence you can write. It is the most transmissible sentence that still says something. Three slots is a workable shape: who you help, what changes for them, and one proof anchor a curious person can check in a single click. The repeater carries the shape and the anchor carries the evidence. That split is the part worth keeping, because it stops asking a third party to vouch for a number on your behalf.
This is a different artifact from a personal mission statement, which works as a decision filter for owners, and the two get confused constantly. A mission statement points inward and only has to survive your own decisions. This sentence points outward and has to survive a stranger with no context and no patience.
Four moves
- Collect the third-person sentences you already have. Message five people who have worked with you and ask, in writing, how they describe what you do when your name comes up. In writing matters, because you need their actual nouns and a phone call hands you a paraphrase filtered through your own ears. Ask for the description rather than the verdict: “when my name comes up, what do you say I do?” is a different question from “what did you think of the work?” and it produces different replies.
- Find the common noun, then fill the three slots. Put the five replies side by side and look for the word more than one person reached for without being prompted. That word is not your positioning yet, but it is the raw material, and it has the one property your own draft never has: somebody else said it first. Build the sentence around it, in three parts, and stop there.
- Run the substitution test. Put a direct competitor’s name at the front of your finished sentence and read it again. If it is still true, you have written a category rather than a claim, and no third party has any reason to produce your name instead of theirs. This catches the sentence that feels sharp and carries nothing, which is the most common failure and the hardest to see in your own copy.
- Put the same sentence in the three places a stranger checks, and make them agree. Your LinkedIn profile, your site, your email signature. Somebody who hears the sentence and then finds three different versions of it has learned something about you that no amount of copy will undo. Consistency here is not tidiness, it is the only corroboration a stranger can get without picking up the phone. Once more than one person writes on the company’s behalf, the business needs its own separate document for this, which is the job of brand voice guidelines, in a 3-step SMB playbook.
What it costs to host the proof anchor
For most readers, nothing. The About section on your LinkedIn profile is free and it is where a stranger looks first. If you already have a site, one existing page does the job, and the work is choosing what goes on it rather than buying anywhere to put it.
If you want a standalone page, two options publish their prices openly. Carrd hosts up to three sites free with its own branding, and its paid tiers are billed per year rather than per month: Pro Lite at $9 a year removes the branding, Pro Standard at $19 a year adds custom domains and forms, and Pro Plus runs $49 a year, with a seven-day Pro trial that does not ask for a card (Carrd). Ghost is the heavier choice and only makes sense if you want the anchor attached to an email list you actually intend to write to, starting at $18 a month billed yearly on Starter for up to 1,000 members, then $29 a month on Publisher (Ghost).
Spending more than that at this stage is usually a way of postponing move one. The anchor’s job is evidence, not design: a named client who agreed to be named, a piece of work somebody can look at, a number you measured yourself and can explain if asked. A beautiful page with nothing checkable on it fails in exactly the same way the vague sentence does.
If you want a structured pass at the first-person half once you have the raw material, the Personal Brand Builder prompt on BusinessPrompter.com is filed under Personal Development and promises a comprehensive personal brand strategy document covering unique value, target audience and actionable steps. It is a Pro prompt behind an upgrade wall and the page shows no price. The sequence is worth being firm about: run it after move one, not before. Fed nothing but your own self-description, any prompt of this kind returns your self-description with better rhythm, which is the input you were trying to get away from.
The claim we would defend
We are about to go further than the evidence strictly supports, and it seems fairer to flag that than to let the paragraphs blend into each other.
The useful division of labor here is not the familiar “AI drafts, you approve.” It is that the model belongs on the half you own and stays off the half you do not. The first-person copy is yours: draft it, compress it, generate ten variants, find the one you can say out loud without wincing. The third-person sentence is not yours at all. It is a report about you, held by other people, produced by work you delivered and by their willingness to put their own standing behind repeating it. There is no prompt for that, and any tool sold on manufacturing “authority” or “thought leadership” as an output is offering you write access to somebody else’s opinion, which nobody has to sell.
Which is the argument for using these tools hard on everything else. The evening spent guessing at posts is the resource actually being wasted, and a model genuinely does hand it back. The only question is where the reclaimed hours go. Five real conversations with people who have worked with you will move the sentence. Another forty drafts will not, however good they get.
Frequently asked questions
Do I need this if all my work already comes from referrals?
Referrals are the thing this is about. A referral is somebody repeating a sentence about you, so if the work arrives that way, the sentence already exists and you have simply never seen it. Move one shows you what is currently being said, which is worth knowing whether or not you change a word afterward.
What if the five replies describe my work five different ways?
That is a result rather than a failure. Five different descriptions mean no sentence is traveling at the moment, and every client you have won arrived through effort you had to spend in person rather than through anything that repeats itself. Pick the description closest to the work you want more of, build the sentence there, and watch whether it starts coming back.
Can I just have an AI write the sentence for me?
It can write the draft and it will write a good one. What it cannot supply is the input that makes the sentence true, which is how other people already describe you, and it will not warn you that the draft is unrepeatable, because agreeing is what these systems do. Hand it the five replies as raw material and it becomes genuinely useful. Hand it only your own bio and it hands the bio back with better rhythm.
How long before this changes anything?
There is no honest schedule, because the sentence travels through other people’s conversations on their timetable rather than yours. What you can check quickly is narrower and more useful: whether the sentence comes back to you. The day somebody you never told repeats your own phrasing at you, it has started moving.
Move one takes about two minutes to send and the replies will not be comfortable reading. Some will be vague. At least one will describe work you stopped doing three years ago. That is the point of it. Read the five with one question in mind, and it is not whether they are flattering. It is whether any two of them reached for the same noun.
