Beyond the Profile, the Presence

When LinkedIn invents four employees for me

LinkedIn says Anara has five employees.

I work alone.

The other four attached themselves to my company without asking, and nobody — not LinkedIn, not me — can stop them.

Anyone can declare themselves an employee of a company on the platform. Nobody checks. The algorithm doesn’t care. My four phantom colleagues form, in the eyes of Sales Navigator, a small coherent structure alongside me: five people with a digital presence, aggregated skills, an organisational footprint. For a recruiter targeting “consultancies with fewer than ten employees in recruiting advisory,” I’m in. For a salesperson prospecting structures “comparable in size to their largest independent client,” I’m in. For a journalist compiling sector headcount figures, I’m in.

Anara is not five people. Anara is not two. Anara is one — mine.

And yet Anara exists as five, in the directory.

That is the heart of what this piece is about. What the tool calls “a company” is not what reality calls “a company.” What the tool calls “a profile” is not what the profession calls “a person.” And that gap — invisible, constant, massive — has been shaping eight years of recruiting and sourcing.

The directory and the register

We conflate two things that, though close, differ fundamentally in nature.

One is a directory. Someone wrote their name, their job, their company, their intentions in it. The directory gathers what people chose to declare about themselves. It holds together through their goodwill — and through the platform’s willingness not to contradict them too often.

The other is a register — the realm of the Act and the fact. The register records a validated event, generally verified by a third party. Information there is objective. The register holds through the truth of facts, whether we like it or not. Someone lived there, was born there, acted there, left traces. It does not confuse itself with the intentions of those it records — and that is precisely why it can serve as proof.

LinkedIn is the first.

The entire profession pretends otherwise.

That pretence is not naivety. It’s convenience. It’s more comfortable to treat the declaration as proof than to go find the proof itself. A keyword typed into a filter gives you five hundred candidates in three seconds. Going to verify which of those five hundred candidates actually does what they claim to do would take five hundred hours.

So we collectively chose to conflate. We called “sourcing” the act of consulting the directory, when we should have kept calling “sourcing” the act of reading the register. The shift happened quietly. And since then, the profession has been manufacturing pipelines of declarations: five hundred candidates who wrote something, five of whom actually did it. We call it productivity. It is substitution.

The proof lives in the numbers LinkedIn communicates about itself — and which do not add up.

On France, Sales Navigator claims thirty-seven million profiles. On the same France, LinkedIn Recruiter claims thirty-four. Two products from the same publisher, drawn from the same base, produce two numbers that differ by three million — roughly the population of Brittany. And both numbers, thirty-seven million as much as thirty-four, far exceed France’s active workforce, which stands at thirty-one million. We are therefore searching for candidates in a base that contains more than exist — retirees, first-year students, vanity profiles, duplicates, twelve-year-old fake accounts like the one I’ve been maintaining since 2013.

Let’s go further. Of those thirty-seven million profiles shown by Sales Navigator, twenty-one million have no documented seniority filter. That’s almost half the base. Which means that the moment you activate that filter — one of the most commonly used in strategic sourcing — you only see half the directory, and you know nothing about the other half. For the tool, you’ve refined your search. For reality, you’ve amputated your vision.

And in the meantime, one account in five disappears every year — one hundred and seventeen million accounts deleted by LinkedIn between January and June 2025. The filters continue to present their stock as if nothing has changed.

A civil registry does not contradict itself. It does not lose twenty percent of its registrants per year without knowing where they went, and it does not pretend to count people it no longer counts.

The tool is powerful because it has the data. But that data is not the truth. It’s a self-reported directory, not a civil register.

Why the LinkedIn illusion persists

Why does this pretence endure? Because it serves us. Four reasons — all comfortable.

The first is speed. A keyword typed into a filter produces a list in three seconds. Mapping an ecosystem, interrogating a community, reading behavioural traces takes hours. Speed has become the quality criterion of sourcing, when it says strictly nothing about its relevance. We confused producing a result with producing the right result. The keyword performs a core sample: it extracts a specimen from one precise spot in the ground. Market Mapping performs a scan: it reveals the structure of the entire subsoil. Both gestures produce images — one shows where you looked, the other shows what is there.

The second is the illusion of objectivity. Filters have the appearance of numbers and the coldness of code. They suggest that a decision made from a Boolean query is better than one made from intuition. Bourdieu had already said it for academic and professional judgements: the more a tool conceals its own subjectivity, the more it allows that subjectivity to operate without scrutiny. A seniority filter that amputates twenty-one million profiles without saying so is not a neutral filter — it has made a decision on your behalf, and you let it because it wore a technical badge.

The third is the business model. LinkedIn never promised to be a register. LinkedIn promised to be a social network, and a social network lives off the time you spend on it — not off the efficiency with which you find what you’re looking for. If you found your candidates in ten minutes, you’d close the tab. If you spent two hours exploring uncertain profiles, comparing, returning, verifying — you’d consume advertising and renew your licence. The directory is not designed to serve you; it’s designed to retain you. Conflating the two is projecting a moral purpose onto LinkedIn’s product that it never claimed.

The fourth is generative AI. Since 2023, LinkedIn profiles have converged. All data candidates are passionate, data-driven, impact-oriented. All salespeople are resilient and results-driven. It’s not that candidates agreed on this — it’s that the same language model suggested the same words to all of them. The candidate didn’t lie; they aligned. And the directory, which already couldn’t distinguish sincere declarations from prepared ones, can no longer distinguish human-written profiles from generated ones. Profiles are no longer written by people — they’re written by the tool that reads them, then re-read by another tool that looks just like it.

The more profiles look alike, the more talent becomes invisible.

That is why the illusion holds. It’s fast, it appears neutral, it serves the platform’s owner, and it self-reinforces through the very tools that were supposed to transcend it. As long as we don’t name these four pillars, we don’t see that they form an edifice — and an edifice, by definition, can be dismantled.

The fault line

What we lost by conflating the directory and the register is not a tool. It’s a way of searching.

Two ways of seeing people coexist in a sourcer’s work, and conflating them means speaking two languages without knowing which one you’re using at any given moment.

The first is that of declaration. It gathers everything someone chose to say about themselves: their title, their company, their skills, their career path, the adjective they stuck in front of their name. It’s voluntary, controlled, crafted. The LinkedIn profile is the typical object of this way of seeing. So is the CV. And so is the client brief, when left in its original form.

The second is that of presence. It gathers everything someone leaves without thinking: their signature in a GitHub commit, their attendance at a meetup, their question posted on Stack Overflow, their name appearing three times on a list of registrants, the Discord community they’ve been running for eighteen months. It’s involuntary, repeated, behavioural. You don’t write it to sell yourself — you leave it because you were there.

The difference between the two is not a question of quantity. It’s a question of nature. The declaration says “here is who I want people to think I am.” The presence says “here is where I was, here is what I did, here is what holds me.” The first is a narrative; the second is a trace. A narrative can lie, can be optimised, can be written by an AI. A trace doesn’t have those freedoms — it has already happened.

Here is what the fault line looks like in practice.

Declaration side	Presence side
The title you display	The projects you ran
The profile you wrote	The community you frequented
The skills you tick	The questions you asked publicly
The “Data Scientist” filter	The active Kaggle account from five years ago
The keyword “Java”	GitHub commits over the past eighteen months
Boolean on Sales Navigator	Repeated attendance at a Java meetup
Static, self-reported	Dynamic, involuntary
A possible lie	A fact that already occurred
Directory	Register

It’s not that one is good and the other bad. It’s that one is fast and the other is true.

The same person exists on both sides — but appears differently in each. A developer’s profile may omit their mastery of Kafka. Their commit from last March cannot omit it. A CRM Manager’s profile may not mention Veeva. Their employer has been a Veeva client for seven years. A Data Science expert’s profile may be frozen at the position they held three years ago. Their presence on arXiv and in two technical communities never is.

The work environment shapes skills. That sentence seems simple — yet it says everything: if you know which ecosystem someone operates in, you know what they can do, even if they haven’t written it. The environment is more reliable than the declaration, because it leaves traces without asking permission.

Two years ago at the Sourcing Summit Europe, I told the story of Thomas. An explorer who goes to find a treasure on an island using an incomplete map, with only a few crosses marked. Thomas, conscientious, follows the crosses. He finds nothing. One evening, exhausted, he meets Alfred — an old fisherman who has lived on the island forever. Alfred looks at the map without impatience and says: “Stop following the crosses. Look at the island.” Thomas starts again, and finds what he had been looking for all along — in the folds, the margins, the territories the map had omitted because they were not listed.

The crosses are the keywords and the filters. The whole island is the presence — with its communities, its behaviours, its traces. Thomas is the sourcer who believed that having points was enough to have a territory. Alfred is the one who teaches him that a territory is not deduced from its points — it’s the inverse.

Master the map, master the market.

The map is not the sum of the crosses: the map is the knowledge of the territory. And the territory is never reached through the directory — it is reached through presence.

The territories of presence

Presence is not an abstraction. It leaves traces, and those traces have addresses.

There is GitHub, where a developer deposits code every day that no one can forge. There is Stack Overflow, where an engineer asks questions that reveal precisely what they’re trying to master. There is Kaggle, where a data scientist participates in competitions that rank them beyond any declaration. There is Reddit, Discord, arXiv, thematic Slack communities, professional associations, trade unions. Each of these platforms is a partial register — a place where presence is proven without being declared.

These territories are not all equally easy to study. Some offer simple technical access; others require dedicated tools. The one I’ve tooled most is called Meetup.

Meetup organises events around shared interests — from Java to baking, from agile to astronomy. People sign up, show up or don’t, come back or slip away. At the end of the year, what remains of a person on Meetup is not what they declared — it’s what they did. Their signature is in their attendance frequency. That frequency has become readable since I built a dedicated toolchain, hosted on freesourcingtools.com, which I run with Guillaume Alexandre, who has built an equivalent tool for arXiv. Every territory awaits its tools.

Three signals are enough to read what it gives you.

The first is engagement. The person comes back to a second event on the same subject. They’re curious. The subject matters to them.

The second is consistency. The person comes back to a third event, a fifth, a tenth. They’re not passing through. The subject is embedded in their practice.

The third is belonging. The person ends up co-organising, presenting, helping with logistics. They no longer consume the community — they make it live.

These three signals — curiosity, consistency, belonging — appear in no LinkedIn profile. And they’re not specific to Meetup: they’re the reading grid that will apply tomorrow to GitHub, Stack Overflow, to any platform where presence accumulates through repetition. Meetup served as the school because it was the first to be technically accessible. The other territories simply await their tools.

A technical rule even allows you to move between them: the same username often travels across platforms. Testing a developer’s Twitter handle against GitHub’s public interface often surfaces their account — and sometimes their email. In the batches I tested at the time, between a third and two-fifths of developers kept the same username across both platforms. For those, you move from declaration to trace in seconds.

This presence can also be inferred. A few years ago, I was looking for a BI application manager who knew Teradata — a rare database technology. The LinkedIn keyword search for “Teradata” gave me eight profiles. On its own, insufficient. But I knew that Bouygues Telecom, SFR, Orange, and a few other large operators used Teradata in their BI architecture — I’d learned this by reading the job postings those companies published. So I redid the search differently: “BI application manager” in the title, and in the company filter a list of known clients of the Teradata vendor. The number of profiles multiplied. None of them had written “Teradata” on their profile. All of them worked with it — because their environment required that mastery.

Eight visible profiles. Several dozen real ones. The difference is not a margin of error. It’s the full set of ideal candidates, hiding exactly where nobody thought to look.

Presence reveals what the profile conceals.

That’s why I now spend more time mapping the territories where people breathe than querying the directory where they’re supposed to have declared themselves. Both gestures cost the same time — but they don’t give access to the same people.

The three tempos

If sourcing can be defined as the act of turning a person into a candidate, I’ve used the analogy of a waltz to characterise its movement. A dance in three beats — easy to anchor in memory, and rather amusing. But it says something true: there are no more than three gestures in a sourcer’s work, and these three gestures always follow in the same order.

The first is targeting. It consists of defining where the people you’re looking for exist — before looking for who they are. This is the phase where you map the ecosystem: companies that are clients of a particular vendor, trade unions in a sector, associations that unite a profession, trade shows that federate a community, market studies that list the players. You haven’t typed a single keyword yet. You’re reading the territory. When Thomas stops in front of Alfred, this is the gesture Alfred teaches him.

The second is identification. Once the territory is circumscribed, you search for the people who live there. This is where you use LinkedIn filters, Boolean searches, Google X-Ray, community platforms. But you no longer use them instead of the method — you use them inside it. The same filter that, used alone, gave eight profiles, becomes the final link in a process that gives several dozen. The difference is not in the tool. It’s in what you did before opening it.

The third is approach. Once people are identified, you must convert them — give them a reason to talk, to meet, to reply. This is the longest gesture, the most human, the most exposed to failure. It doesn’t belong to the directory or the register. It belongs to the profession itself, in its most ancient form.

These three tempos — targeting, identification, approach — have been my method since 2018, at a time when I had neither a toolchain, nor a useful filter, nor artificial intelligence to hold my hand. The method has not changed since. What has changed are the tools it mobilises at each of its tempos. The order of the gestures has remained the same — because it doesn’t depend on technology. It depends on what a sourcer is looking for, and on how a human allows themselves to be found.

A concrete demonstration I’ve been showing to the sourcers I train for years.

A client asks for a CRM Omnichannel Manager who knows Veeva, widely used in the pharmaceutical industry. I start with the lazy method — the one that skips the first tempo: I type “Veeva” as a keyword on LinkedIn, filtering for the pharmaceutical industry, the Paris region, Omnichannel or CRM titles. Twenty-seven profiles. For a strategic hire, that’s a statistical dead end.

I start over from the beginning. First tempo: targeting. I ask ChatGPT, then Perplexity, then the vendor’s official client page, for a list of companies that use Veeva in their tech stack. I get fifty-five companies. Second tempo: identification. I build a Sales Navigator query where I keep the title CRM Omnichannel Manager, the geography, the industry — but I replace the keyword “Veeva” with the list of fifty-five client companies. One hundred and six profiles. After deduplication with the first search: one hundred and seventeen people in total. Multiplied by four. And crucially, none of those profiles had written “Veeva” on their own profile — they use it because their employer is a client.

The method is not new. Twenty-five years ago, a young consultant working on key account recruitment in consumer electronics, my manager used to send me to the FNAC for a store check: I’d note which brands were on the shelf, reconstruct the organisational charts of the manufacturers back at the office, and only then look for who, at each one, was managing the commercial relationship with the retailer. The store check of the consumer has become, with three digital tools, the market mapping of today. The technique changed; the gesture never moved.

There is even a frugal version of this targeting, for markets where no institutional database is available. I call it the fifth element technique. You identify four obvious competitors — the ones everyone in the sector knows — and search for them together in Google, in quotes. Google returns the pages that list all four of them: trade associations, business press, market studies, trade shows. Those pages almost always list others — the fifth competitor, the tenth, sometimes the fiftieth. It’s free, fast, always available. And it’s precisely the first tempo of the method.

The method is stable. The tools pass.

That is what eight years of mapping have taught me: when you learn to hold steady across the three tempos, you stop fearing changes in tools. ChatGPT will tomorrow replace what Google does today. Something else will replace ChatGPT in four years. But targeting that precedes identification that precedes approach — that rhythm is replaced by no artificial intelligence. It is the infrastructure of the profession.

The ethics

Preferring presence over declaration is not just a methodological decision. It’s a decision about how you look at people. And the moment you look differently, you engage something beyond your efficiency.

Reading a profile and reading a trace are not the same gesture. The profile is a voice telling its own story: it chooses its words, tends its shop window, corrects its angles. The trace was not chosen for us. Someone pushed a commit on GitHub because they were debugging at one in the morning — not to be found. Someone came to the Paris Symfony meetup four times because that subject matters to them — not to sell themselves to a recruiter. The trace says something the profile cannot say, and that the profile often does not want to say.

Hence a sentence I’ve been repeating for years, in my articles and in my training sessions: proof is the antidote to the declarative. What has been done says more than what is asserted. The declarative can lie, embellish, deceive itself. Proof cannot — it precedes the narrative drawn from it.

Choosing presence is therefore choosing a form of responsibility. When you stick to the profile, you can always hide behind the filter: LinkedIn surfaced it, the algorithm ranked it, the tool sorted it. The sourcer who learns to read traces can no longer hide behind anyone. They look, they infer, they decide. The tool is no longer the judge; it becomes an instrument again. The data illuminates — but it does not decide.

This asymmetry also changes the relationship to the person being sought. The profile slots candidates into boxes: Head of Data in Paris, eight years’ experience. The trace puts them back in their full thickness: a career path, a curiosity, a practice, a community. We don’t recruit CVs. We bet on stories. And a bet, unlike a filter, commits the one who bets as much as the one bet upon.

This is perhaps what the profession lacks most today: not faster tools, but the conviction that a sourcer is an investigator of the living — not a filter operator. Investigation takes time, it takes doubt, it requires accepting that you can be wrong. Above all, it requires that you take back ownership — of what you’re looking for, of what you see, of what you conclude from it.

The directory relieved us of the need to look. The register compels us to.

Three questions to ask before opening LinkedIn

Where am I going, where am I coming from — and above all, what am I looking for?

One question remains — simple — that a sourcer can ask themselves any Monday morning before opening Recruiter or Sales Navigator: what am I looking for, and where am I going to look for it?

Three gestures are enough to answer it.

None of them requires a new tool.

• Before typing the first keyword: do I know where the people I’m looking for live — or am I waiting for LinkedIn to know that for me?
• In front of a profile: am I reading what they declare, or am I looking for what they leave behind?
• After adding them to my pipeline: am I trusting the filter, or my own judgement?

These three questions don’t protect you from error. They protect you from comfort.

That moment where you confuse the profile with the person,

The declaration with the presence,

The directory with the register.

The profession begins when you stop confusing them.

Stop reading profiles. Learn to read traces.