The personal branding industry runs on a statistic from 2012

The most quoted statistic in personal branding is fourteen years old and was published by a company that sold the exact service it recommends. Eighty-two percent of consumers are more likely to trust a company whose CEO is active on social media. You have seen it this quarter, probably in a deck, possibly in one of mine. It comes from BRANDfog's CEO, Social Media and Leadership Survey, released in March 2012 by a firm selling CEO social media services, and the methodology was never published.

That is not a small embarrassment in a corner of the field. It is load-bearing. An industry whose entire product is trust makes its opening argument with evidence that would be laughed out of a research seminar, and the pattern repeats all the way down. The 561 percent reach figure is a tweet. The 57 percent buying journey figure predates the modern LinkedIn feed. The newest one, that 81 percent of long-form LinkedIn posts are AI-generated, came from a company selling AI detection.

Here's the thing. I am not writing this to score points on other consultants. I have quoted two of these numbers myself, and I described a credential of my own in a way I would now change, which I will come to at the end. The reason to fix it is selfish. Buyers check now, checking costs nothing, and a number you cannot source is the fastest way to tell a careful person that you are not careful.

Where does the 82% CEO trust statistic actually come from?

The findings travelled because they were quotable: 82 percent more likely to trust, usually paired with 77 percent more likely to buy and 79 percent who would prefer to work there. Three round numbers saying exactly what a consultant wants them to say. The survey covered what the release called hundreds of employees, the sample was never broken out, and no methodology was published beside it.

Start with the date, because the date alone should end the conversation. It is fourteen years old and it predates the modern LinkedIn feed entirely. Being visible online meant something so different then that the answers cannot transfer. Asking people in 2012 how they felt about chief executives on social media tells you about 2012, and putting that on slide two of a 2026 proposal is not evidence, it is decoration.

The age is the obvious problem. The authorship is the worse one. A company surveying the market for its own product, releasing the result through a wire service and declining to show its working is not doing research, it is marketing in a research costume. Vendors have always done that. The scandal is that fourteen years on our field still repeats it unchecked, which means most people quoting it have never clicked through to see what they are standing on.

The '82% of consumers are more likely to trust a company whose CEO is active on social media' statistic comes from a March 2012 survey by BRANDfog, a firm that sold CEO social media services, and its methodology was never published.

If you want to argue that executive visibility matters, and I do, argue it from behaviour rather than from a fourteen-year-old survey about feelings. H/advisors Abernathy counted what chief executives actually did.

Read that pairing carefully, because it says something the 2012 stat structurally cannot. Chief executives cut their output by more than a third and got more response for it. That is a finding about restraint, and it is inconvenient for anyone selling a five-posts-a-week retainer, which is probably why you have not seen it in a deck.

What is the 561% employee reach number, really?

The claim is that brand messages reach 561 percent further when shared by employees than through official brand channels. It appears in employee advocacy pitches constantly, usually with no year attached and occasionally with a vague nod towards a report. I spent an afternoon looking for that report. So has everybody else who went looking.

A tweet is not a study. It may once have pointed at one, and MSLGroup was a real agency doing real work, but a claim you cannot open is a claim you cannot defend. Asked what the sample was, which platforms and which year, you have nothing to hand over. You are quoting an eleven-year-old social post to someone deciding whether to trust you with their reputation.

The mutation should worry you more than the missing report. Several 2026 sources now restate the same 561 percent as personal profiles driving more reach than company profiles, which is a different claim entirely. Employee sharing against brand channels and personal profiles against company pages do not share a denominator, and no single study could have produced both. When a number survives intact while the thing it measures changes underneath it, you are not looking at evidence. You are looking at folklore with a decimal point.

A number you cannot open is not evidence. It is a rumour with a decimal point.

Now the honest part, because the direction here is right. Personal profiles do reach further than company pages. LinkedIn's own product design points that way, several datasets agree, and anybody who has run both accounts for a year has felt it. What does not exist is a defensible multiplier. The 5x, the 7x, the claim that company pages get five percent of feed allocation against sixty-five for personal profiles, all of it traces to vendor blogs rather than to LinkedIn or an audited dataset. Keep the direction, drop the multiplier.

1. c. 2011
   CEB buyer researchOrigin of '57% of the buying journey is complete before contacting sales'. A headline average sitting on very wide variance.
2. March 2012
   BRANDfog CEO surveyOrigin of '82% trust a company whose CEO is active on social'. Published by a firm selling CEO social media services. No methodology.
3. January 2015
   MSLGroup tweetOrigin of '561% further reach' for employee-shared content. The underlying report has never been located.
4. July 2026
   Originality.ai studyOrigin of '81% of long-form LinkedIn posts are AI'. Published by a company that sells AI detection, scored by its own detector.

Is 57% of the buying journey over before you talk to sales?

This one has done more damage than the other two combined, because it shaped strategy and not just slides. The origin is CEB research from around 2011 and 2012, and even then it was a headline average sitting on enormous variance between deal sizes and industries. Such an average is not wrong. It is useless as a planning input, the way the average depth of a river is useless when your real question is whether you can walk across it.

Then came the escalation. 57 percent became 70 percent became 80 percent, each figure lifted from a different research house, in a different year, using a different definition of when the journey starts. Stacked in a slide they look like a trend line. They are three measurements of three different things, and putting them in ascending order does not make them a series. It makes them a story someone assembled afterwards because the story sold content.

The 57%, 70% and 80% figures for how much of the buying journey is complete before contacting sales come from different research houses in different years using different definitions, so they do not form a trend.

And the direction is now wrong, which is the part almost nobody has updated for.

That runs against the folk narrative, and 6sense says why. 62 percent of buyers reported economic factors pushing them to engage sellers sooner and 49 percent said financial pressure shortened their cycle. Under pressure, buyers stop browsing quietly and start asking early. So the job of your content is not to replace the first conversation, it is to earn an earlier and better one. If you have spent two years telling clients that buyers drift further away every year, the best current measurement says your compass points backwards.

• Point of first contact69% of the journey61% of the journey
• Research vs seller engagement70/3060/40
• Average buying cycleAbout 11 monthsAbout 10 months
• Direction of the storyBuyers drifting further from salesBuyers engaging sellers earlier

Is 81% of LinkedIn actually written by AI?

This is the newest zombie and the most repeated number in the field right now. Originality.ai published it in July 2026 and it was everywhere inside a week, because it confirms what everyone already feels while scrolling. I suspect the feeling is broadly correct. The number is still not evidence, for three reasons, any one of which would keep it out of your deck.

The first is who published it. Originality.ai sells AI detection software, so a finding that most of LinkedIn is machine-written is an advertisement that costs nothing to produce and earns coverage for free. That is structurally identical to BRANDfog in 2012, fourteen years and a whole technology cycle later.

The second is the sample. It is 5,000 posts pulled from ten pages of search results across ninety topic-and-date searches. That is not the feed. It is the slice of LinkedIn that search engines choose to surface, which by construction over-represents keyword-shaped writing and under-represents the ordinary posting that is most of the platform. To their credit the authors concede the sample is narrower than feed-based studies. Almost nobody repeating the headline mentions that.

The third is the instrument. The posts were scored by the publisher's own detector at a 15 percent AI allowance threshold, and detectors are nowhere near reliable enough to carry a headline this size.

Sit with that second figure. Seven detectors flagged 61.3 percent of TOEFL essays as machine-written and almost none of the native English writing. TOEFL essays are written by people learning to write formal English, and formal second-language English carries exactly what a detector reads as synthetic: careful clause construction, limited idiom, low variance in sentence length, vocabulary chosen for safety rather than colour. LinkedIn's membership is overwhelmingly people writing English as a second or third language. Run an instrument with that bias across that population and there is a real chance you are measuring accent rather than automation.

AI detectors misclassify human writing as machine-written 12 to 26 percent of the time, and one evaluation found 61.3 percent false positives on essays by non-native English writers, which makes any detector-based estimate of AI content on LinkedIn unreliable.

• Human text, leading detectors (low end)12%
• Human text, leading detectors (high end)26%
• TOEFL essays, non-native English writers61.3%

Two footnotes make it worse. The study page discloses that significant AI editing was used on the post announcing the finding. And travelling beside the 81 percent is a companion figure with no source at all: that LinkedIn's authenticity update cuts AI content reach by up to 47 percent, sometimes rendered as 45 percent less engagement. Neither appears in any LinkedIn publication. LinkedIn has published architecture, not reach deltas by content type, so every percentage you read about AI penalties was invented by somebody who needed one.

Which is the frustrating part, because the real evidence is stronger than the invented evidence. A platform that profits from posting volume deleting its own AI writing button is a more powerful fact than any percentage, since it is dated, attributable and impossible to argue with. The fake 47 percent gets you a slide. The July announcement gets you an argument that holds when somebody pushes back.

Who audits the numbers the ghostwriting industry quotes about itself?

Nobody does, and it is worth saying plainly, because the field talks about itself with the confidence of a sector that has a trade association and an annual census. There are more than 200 ghostwriting agencies globally. Demand is growing roughly three times year over year. The executive creator economy grew 34 percent year over year in the first quarter of 2026. Follow any of those to its source and you land on an agency blog citing another agency blog citing a third that has since been deleted.

There is no trade body, census or independent market study of LinkedIn ghostwriting, so every growth figure circulating about the industry traces back to agency blogs citing each other.

Be fair about which half is fabricated. The price ranges quoted, roughly one and a half to four thousand dollars for a freelancer, four to eight for a studio, eight to fifteen for an agency, are plausible as aggregated anecdote and match what practitioners tell each other privately. The growth rates are a different species, because a growth rate needs somebody to have counted the same population twice, and nobody has counted it once.

Two errors in one sentence, and both happened by copying rather than by lying. A number gets repeated, someone rounds it, someone else credits the bigger and more familiar brand because that feels authoritative, and within a year the circulating version has the wrong sample and the wrong author. This is what happens to any number that travels without its paperwork.

Does quoting a bad statistic actually cost you anything?

Let me put the strongest version of the objection, because I have heard it from people I respect. Nobody has ever lost a deal because slide four cited a 2012 survey. Clients do not check. Prospects want a confident story and a reason to say yes, and the consultant who stops mid-pitch to explain sampling is the one who does not close. For most of the last decade that was simply true.

It stopped being true recently, and not because of an outbreak of rigour. It is a change in cost. 6sense found that 94 percent of B2B buyers used large language models during the buying process to summarise reviews or analyse data. Checking a claim used to cost twenty minutes of somebody's afternoon, so nobody bothered. It now costs one sentence typed by an analyst already sitting in a chat window. Your deck is being fact-checked whether or not anyone in the room announces it.

There is a second cost that matters more, and it took me years of losing deals I thought had gone well to see it.

Put those findings side by side and the picture changes shape. Most B2B decisions are not made in front of you. They are made in a meeting you were not invited to, by somebody who liked you, arguing your case to colleagues who never met you and are looking for a reason to say no. Everything you publish is ammunition for that one person. An unsourceable statistic is the worst ammunition you can hand them, because when a skeptical CFO types it into a search box and finds a fourteen-year-old vendor survey, it is your advocate who looks careless. You did not lose the argument. You lost the person making it for you.

In B2B buying, thought leadership works as ammunition for an internal advocate arguing your case in a meeting you are not in, which is why an unsourceable statistic damages that advocate more than it damages you.

And then there is the part specific to us. A logistics company that misquotes a statistic loses a citation and nothing else. We sell trust as the actual product, so being wrong about the evidence for trustworthiness is not a flaw in the argument, it is a live demonstration against it. That is why I would rather publish a thinner claim I can defend than a thicker one I cannot.

How do you check a number in under five minutes?

Five questions, in order, and you can run most of them while somebody is still talking. Numbers usually fail on the first or the fourth.

• Who paid for it? If the publisher sells the thing the number recommends, treat it as an advertisement until proven otherwise. BRANDfog sold CEO social media services. Originality.ai sells AI detection.
• What is the sample, and is it what people did or what people said? Observed behaviour beats stated intention, and most famous numbers in this field are stated intention from a few hundred people.
• What is the date, and did the thing it describes still exist in that form? A 2012 survey about executives on social media predates the modern LinkedIn feed entirely.
• Can you reach the primary document? If three clicks do not land on a PDF, a paper or the publisher's own newsroom, you do not have a source, you have a rumour with a good reputation.
• Has the claim changed while the number stayed the same? That is the clearest sign of a zombie. 561 percent began as employee sharing against brand channels and is now restated as personal profiles against company pages.

When a number fails, resist the reflex to hunt for a replacement percentage. The honest sentence is usually stronger than the statistic it replaces. Saying that personal profiles reach further than company pages, that LinkedIn's product design points that way, and that nobody outside the vendors has published a multiplier you would defend, makes you the only adult in the meeting. It also does what a borrowed number never can, which is show the judgment the client is actually buying. They are paying to find out whether you can tell a fact from a claim, because that is the skill they rely on when you write in their name.

	Dated	Method published	Independent of seller	Measures behaviour
'82% trust a CEO on social'	2012	×	×	×
'561% further reach'	2015 tweet	×	×	×
'57% of the buying journey'	c. 2011	−	✓	×
'81% of LinkedIn is AI'	2026	−	×	−
'47% reach penalty for AI posts'	no date	×	×	×
'Ghostwriting demand growing 3x'	no date	×	×	×

What should you cite instead?

The replacements exist, they are free to read, and most make better arguments than the zombies they replace. Start with the one that dismantles the objection every founder actually raises, which is that the feed is saturated and there is no point joining now.

Sit with that, because it takes apart the most common reason people give for staying quiet. At the executive level the feed is not saturating, it is emptying. Roughly half the chief executives visible in early 2023 are not visible now. Whatever you think of the quality of what remains, the room is getting less crowded, and scarcity is a far better argument for visibility than borrowed trust from 2012 ever was. It also survives a hostile question, which the 82 percent has never once done.

The same dataset holds a finding almost nobody acts on. Companies with the most active VP and Director layer see four times the employee posting of those with the least active, against only 2.6 times for the same C-suite comparison. If you can make one layer visible, do not start at the very top.

The second replacement answers the question every founder eventually asks, which is whether any of this reaches a buyer. 6sense found 94 percent of buying groups had ranked their preferred vendors before speaking to a single seller, then bought from that pre-contact favourite roughly 77 to 80 percent of the time. Read that as an instruction rather than a statistic. The shortlist is written before you are in the conversation, so the only question is whether you existed, in a form somebody could evaluate, on the day it got written.

• 13-15%CXO-level leaders posting in a given month, down from 25% in early 2023
• 94%Buying groups that ranked preferred vendors before speaking to any seller
• 79%Decision-makers more likely to advocate for you in an RFP if you publish well

One caution, because I would be doing the thing I am criticising if I skipped it. None of these publishers are disinterested. Each works inside the business its own research describes, and each wants visibility to look valuable. The difference is not purity, it is disclosure. They give you the sample size, the dates, the method and the definitions, so you can argue with them. BRANDfog did not, which is why you cannot. A source you can pick a fight with is a source.

What happened when I ran this test on my own bio?

I owe you the same audit I have just run on everybody else, and it does not come out clean. For a while I described a Forbes appearance as being named to a list of top personal branding experts. That is a small upgrade on the truth, and small is how these things always start.

So what is true? I am named in that January 2025 Forbes piece, and quoted again in Forbes on 13 January 2026 on why leaders cannot fake authenticity. Both are real and checkable in one click, and the accurate phrasing is that I was featured among top global personal branding experts in Forbes in January 2025. That is also the better line, because a claim a journalist can verify instantly carries more weight than a bigger one they cannot.

Here is how a bio line drifts, and it is exactly how BRANDfog's survey became a fact. Nobody sits down and decides to lie. You write a slightly generous description of a real thing, a client repeats it in an introduction, an event page adds an adjective, and eight months later the inflated version arrives back in a stranger's email and you no longer remember the original. The credential got worn smooth by handling, the same way the 82 percent did. So the discipline is not honesty in the moment, since almost everybody passes that test. The discipline is going back to check what you said last year.

You do not have to be right about everything. You have to be checkable about everything you claim.

The uncomfortable part is how cheap the fix is and how rarely anyone bothers. Every replacement here is public, dated and free, and all of them are linked at the bottom. The work is not research. The work is giving up the more exciting number for the one you can defend, which is a decision about character rather than method, and that is why so few people make it.

So start today, with one thing. Take the statistic you use most often when you explain why your work matters, the one that would go on slide two, and spend ten minutes finding its primary source. If you land on a PDF with a sample size and a date, you are fine, and you now know something most of your competitors do not know about their own pitch. If you land on a blog citing another blog, you have learned something more useful than a statistic: the strongest-sounding part of your argument was quietly the weakest.

Then ask the harder question, the one I had to ask about my own bio and did not enjoy. If a buyer checked every claim you make about yourself this week, which one would you want them to get to first?

Questions I get asked about this

Where does the '82% of consumers trust a company whose CEO is on social media' statistic come from?

BRANDfog's 2012 CEO, Social Media and Leadership Survey, released in March 2012 by a firm that sold CEO social media services. It surveyed what the release called hundreds of employees and published no methodology. It is fourteen years old and predates the modern LinkedIn feed entirely, so it should not be cited as current evidence.

Is the 561% employee reach statistic real?

The number traces to a January 2015 post on MSLGroup's Twitter account, and the underlying report has never been located. Its year is unconfirmed even by the vendors citing it, and it is now being restated as a claim about personal profiles versus company pages, which is a different comparison wearing the same number. The direction, that personal profiles reach further than company pages, is well supported. The multiplier is not.

Is 57% of the buying journey complete before buyers contact sales?

That figure comes from CEB research around 2011 and 2012, and the popular 57 to 70 to 80 percent escalation stitches together different studies with different definitions. The best current measurement runs the other way: 6sense's 2025 Buyer Experience Report, published 12 November 2025, moved point of first contact down from 69 percent to 61 percent of the journey, with 62 percent of buyers saying economic factors made them engage sellers earlier.

Is 81% of LinkedIn content AI-generated?

That figure comes from Originality.ai, a company that sells AI detection, scored by its own detector on 5,000 search-discoverable posts rather than a sample of the feed. AI detectors misclassify human writing 12 to 26 percent of the time, and one evaluation found 61.3 percent false positives on TOEFL essays by non-native English speakers against near-zero on native English writing. On a platform where most members write English as a second language, treat the figure as unproven.

What personal branding statistics can I actually cite in 2026?

Tribal Impact's behavioural benchmark of 860 B2B companies published 20 July 2026, 6sense's 2025 Buyer Experience Report published 12 November 2025, the 2025 Edelman and LinkedIn Thought Leadership Impact Report published 8 July 2025, and H/advisors Abernathy's Social in the C-Suite report published 18 December 2025. All four publish their sample size, date and method, and two of them measure behaviour rather than opinion.

Sources

1. BRANDfog 2012 CEO, Social Media & Leadership Survey, BRANDfog, March 2012.
2. BRANDfog Survey Reveals CEO Engagement In Social Media, Business Wire, 12 March 2012.
3. MSLGroup post, origin of the 561% employee reach figure, MSLGroup (Twitter/X), January 2015.
4. The B2B buying decision process: challenging the 57% myth, Inflexion-Point, Undated, accessed 2026.
5. The Timeline for Influencing B2B Buyers Is Shrinking: Insights From 6sense's 2025 Buyer Experience Report, 6sense, 12 November 2025.
6. The 2026 Employee Advocacy Behavioural Benchmark Study, Tribal Impact, 20 July 2026.
7. 2025 B2B Thought Leadership Impact Report: Invisible Influence, Edelman and LinkedIn, 8 July 2025.
8. Social In The C-Suite Report: Four questions CEOs face regarding their use of social media, H/advisors Abernathy, 18 December 2025.
9. LinkedIn AI Content Study: 81% of Long-Form Posts Are Likely AI, Originality.ai, July 2026.
10. Evaluating the accuracy and reliability of AI content detectors in academic contexts, International Journal for Educational Integrity (Springer), 2026.
11. AI Detection and assessment: an update for 2025, Jisc National Centre for AI, 24 June 2025.
12. LinkedIn adds a button to report AI-generated slop, TechCrunch, 30 July 2026.
13. Engineering the next generation of LinkedIn's Feed, LinkedIn Engineering, 12 March 2026.
14. LinkedIn Organic Benchmarks 2026, Socialinsider, 2026.
15. 19 Essential Personal Branding Tips From Top Global Experts, Forbes (William Arruda, Senior Contributor), 5 January 2025.
16. Why Leaders Can't Fake Authenticity In The Age Of Social Media, Forbes (William Arruda, Senior Contributor), 13 January 2026.

Every figure above is linked to its original source. Where I am predicting rather than reporting, I say so in the panels marked My take.