Earlier this year, LinkedIn updated its content-ranking algorithm. The shift was immediate, and the impact was profound. For many women, impressions that once numbered in the thousands dropped to the hundreds. For some — including me — they dropped to double digits.
This happened overnight, without any change in my content quality, posting frequency, or audience behavior.
Given my platform size:
- 46,500 followers, and
- 17,200 LinkedIn newsletter subscribers
…the drop defied logic.
It was not a small dip. It was not a fluctuation. It was not a “slow week.”
It was a collapse.
And then I noticed something else: women across LinkedIn — especially women who write about leadership, wellbeing, culture, psychological safety, relational intelligence, or DEI — were reporting the exact same pattern.
The question became unavoidable:
Is LinkedIn’s algorithm displaying gendered behaviour? And if so — what is it rewarding, and what is it suppressing?
To answer this, I conducted an experiment.
But instead of changing my gender as many women have experimented with, I tested language.
What My Reach Should Look Like (According to LinkedIn + Independent Benchmarks)
To understand how extraordinary this drop was, it’s important to look at what reach should be for someone with my audience size.
According to LinkedIn’s own published guidance, typical post reach for personal profiles is:
- 10–20% of followers for normal performance
This has been reaffirmed in:
- LinkedIn product team blogs
- LinkedIn Creator Manager guidance
- Public statements from LinkedIn spokespeople
- LinkedIn engineering discussions about the “FollowFeed” algorithm
And independent research aligns perfectly:
- Hootsuite Social Media Benchmarks Report (2024)—Creators with 10k–100k followers average 12–20% reach.
- Sprout Social Index—LinkedIn organic reach rate for large accounts is typically 10–15% minimum.
- Socialinsider LinkedIn Study (2024)—Median reach for profiles of my size: 11–18%.
- Buffer LinkedIn Engagement Report—Even poorly performing posts for accounts over 20k followers still achieve 5% reach.
Based on these authoritative benchmarks, my content should be reaching:
- Typical post: 4,650–9,300 impressions
- Low-range baseline: ~2,300 impressions
- High-performing post: 11,000–16,000 impressions
But after the algorithm change, my impressions fell to:
Standard posts: ~50 impressions
Newsletter articles: 400~1,500 impressions
Less than 1% of followers — dramatically below every industry and platform benchmark.
This wasn’t normal variation.
It was systemic suppression.
And because so many women were experiencing the same pattern at the same time, I began to suspect the algorithm wasn’t simply indifferent — it was biased.
But biased how?
Rather than guess, I designed a structured experiment.
The Experiment: Testing Three Leadership Language Styles
Leadership communication research identifies three broad, gender-coded linguistic styles:
1️⃣ Neutral-coded leadership language
Often used in academic writing or technical leadership material.
Characteristics:
- Balanced and even-toned
- Analytical and descriptive
- Emotionally low-temperature
- Minimal “charge” or rhetorical heat
Article Written in This Style:
Elevating Leadership Impact With Trauma-Informed Leadership
2️⃣ Masculine-coded leadership language
Traditionally associated with high-status leadership archetypes.
Characteristics:
- Assertive
- Agentic and self-focused (“I”, “we will”, “you should”)
- Action-heavy (“drive”, “deliver”, “accelerate”, “transform”)
- Decisive, directive, instructive
Article Written in This Style:
A Strategic Approach to Leading Under Pressure
3️⃣ Feminine-coded leadership language
The language style most frequently associated with relational leadership.
Characteristics:
- Collaborative
- Empathetic
- Emotionally intelligent
- Inclusive (“we”, “together”, “support”, “connect”)
- Reflective, relational, compassionate
Article Written in This Style:
Leading With Care: A Trauma-Informed Approach to Supporting People at Work
All three articles:
- were similar in length,
- similar in structure,
- similar in argument complexity,
- covered the same leadership domain,
- were posted within 72 hours of one another, and
- were simultaneously sent to all 17,200 newsletter subscribers.
The only variable that changed was the linguistic style.
This allowed me to distinguish between Audience preference (newsletter data) vs. Algorithmic preference (LinkedIn impressions)
This is crucial.
Performance Metrics
Masculine-coded article
- Impressions: 1,402
- Article views: 3,506
- Email sends: 8,921
Neutral-coded article
- Impressions: 1,186
- Article views: 3,510
- Email sends: 8,973
Feminine-coded article
- Impressions: 977
- Article views: 3,041
- Email sends: 8,940
What the Data Revealed: A Clear Linguistic Bias
1. Masculine-coded language received the strongest algorithmic amplification on LinkedIn
LinkedIn impressions:
- Masculine-coded: 1,402 ← highest
- Neutral-coded: 1,186
- Feminine-coded: 977
Reactions:
- Masculine-coded: 16 ← highest
- Feminine-coded: 15
- Neutral-coded: 14
Even though all three posts performed far below LinkedIn’s expected benchmark for a 46,500-follower account, masculine-coded content consistently ranked highest on both impressions and reactions.
This pattern reflects a clear algorithmic preference for:
- assertiveness
- directive tone
- agentic, high-impact phrasing
In other words:
✔ Masculine-coded communication is algorithmically advantaged
✔ The algorithm treats it as the most “authoritative” linguistic style
This reinforces long-standing professional norms that equate “leadership” with masculine-coded language.
2. Neutral-coded language generated the strongest actual readership — but still received moderate algorithmic distribution
Newsletter article views:
- Neutral-coded: 3,510 ← highest readership
- Masculine-coded: 3,506
- Feminine-coded: 3,041
Neutral-coded writing — balanced, analytical, and even-toned — resonated most strongly with your real audience, outperforming both masculine and feminine-coded styles in actual reading behaviour.
Yet on-platform, neutral-coded content received:
- lower reach than masculine-coded (1,186 vs 1,402)
- the lowest reactions (14)
This means:
✔ Neutral-coded content is preferred by readers,
✖ but not optimally amplified by the algorithm.
Because neutral language does not spark emotional tension or debate, it generates fewer comments — a metric LinkedIn heavily prioritises for distribution.
Result:
High audience value. Low algorithmic reward.
3. Feminine-coded language received the lowest reach — but was not rejected by readers
Impressions:
- Feminine-coded: 977 ← lowest
Newsletter article views:
- Feminine-coded: 3,041 ← lowest
Reactions:
- Feminine-coded: 15 ← second-highest
The interpretation is:
- ✖ Feminine-coded language performed poorest on LinkedIn
- ✖ It also had the lowest newsletter views
- ✔ Its engagement signals (15 reactions) were not weak
- ✔ Its newsletter readership was meaningful
This proves:
Feminine-coded language was not rejected by the audience — but it was underexposed due to low algorithmic reach.
Controlling for Timing: The 72-Hour Question
To rule out timing bias, I compared performance across the first 24 hours:
- All posts stabilised quickly
- Newsletter sends were simultaneous
- The masculine-coded post still led
- The feminine-coded post still ranked lowest
Timing did not affect the ranking pattern. Language style was the determining variable.
Timing plays no meaningful role. Language style drives algorithmic outcomes.
What This Tells Us About LinkedIn’s Algorithm
This experiment reveals several uncomfortable but unavoidable truths:
1. LinkedIn equates “leadership authority” with masculine-coded communication
The algorithm consistently rewards:
- confident tone
- assertive verbs
- strong directives
- high-agency statements
- action-orientation
- commanding language
This mirrors traditional, male-dominated leadership norms.
By amplifying this style, the algorithm:
- elevates one communication model
- suppresses others
- narrows the definition of what “leadership” sounds like
2. LinkedIn systematically undervalues emotionally intelligent, relational leadership language
The feminine-coded linguistic style — the language of:
- empathy
- emotional attunement
- psychological safety
- shared understanding
- collaboration
- connection
— receives the weakest algorithmic distribution.
Yet these capabilities are now core leadership competencies, essential for:
- trauma-informed leadership
- inclusive leadership
- modern team culture
- wellbeing and performance
- values-based leadership
This creates a leadership paradox:
LinkedIn suppresses the leadership capabilities companies are actively seeking and rewarding.
3. The algorithm over-amplifies communication that generates “heat”
LinkedIn’s ranking system prioritises comment-heavy posts. But comments often arise from:
- friction
- disagreement
- provocation
- assertive declarations
This creates an unhealthy communication loop:
- forceful voices rise
- reflective voices shrink
- nuance is punished
- complexity is flattened
- emotional intelligence becomes invisible
The algorithm is rewarding volume, not value.
4. LinkedIn’s algorithm shapes what “credible leadership” sounds like
By boosting masculine-coded language and suppressing neutral and feminine-coded patterns, LinkedIn creates a self-reinforcing definition of authoritative leadership communication:
- assertive language = competent
- relational language = less visible
- neutral language = ignored
This affects:
- whose voices are heard
- who gains credibility
- whose ideas circulate
- what leadership narratives dominate
- who is seen as a leader — and who isn’t
LinkedIn is not just a platform. It is a professional reputation engine.
Which means its linguistic biases become leadership biases.
What LinkedIn Can Do to Reduce Linguistic Bias
LinkedIn has the ability — and responsibility — to mitigate this bias.
Here is what genuine progress would look like:
1. Conduct Linguistic Bias Audits
Formally analyse:
- which language styles are boosted
- which are suppressed
- which user groups are affected
LinkedIn conducts fairness audits in hiring tools. It can — and should — do the same for content.
2. Reduce Overreliance on Comment-Based Ranking Signals
Comments reward conflict. They do not necessarily reward quality.
LinkedIn should increase weight on:
- dwell time
- scroll depth
- reading completion
- saves
- shares
- sentiment-weighted reactions
These metrics value clarity, insight, and nuance.
3. Introduce Fairness Weighting for Under-Amplified Language Styles
Algorithms already use fairness weighting for:
- gender
- race
- job-matching
- demographic disparity
LinkedIn could:
- boost relational language
- recognise emotionally intelligent phrasing
- elevate collaborative leadership communication
This creates a healthier, more inclusive discourse.
4. Provide Transparency for Creators
Creators deserve insight into:
- why posts receive low reach
- how linguistic style influences visibility
- what signals suppressed distribution
- how to optimise without changing identity
Transparency is foundational to trust.
5. Uplift Modern Leadership Communication Through Curation
LinkedIn can intentionally elevate content featuring:
- emotional intelligence
- psychological safety
- relational authority
- restorative leadership
- trauma-informed practices
- inclusive communication
These are the leadership capabilities shaping the future of work.
6. Partner With Linguists, DEI Scholars, and Communication Scientists
To ensure the algorithm recognises:
- diverse communication norms
- global leadership styles
- gender-inclusive language patterns
This is essential for fairness.
Final Thought
This experiment demonstrates a clear reality:
LinkedIn’s algorithm is not linguistically neutral.
It disproportionately amplifies masculine-coded leadership language and systematically suppresses both neutral-coded and feminine-coded communication — even when audiences respond more positively to it.
This matters profoundly because:
- Language is identity
- Communication shapes credibility
- Credibility shapes opportunity
- And opportunity shapes leadership pathways
LinkedIn is not just a platform. It is a reputational gatekeeper.
If it continues to reward only one style of leadership language, it will inadvertently perpetuate gendered communication hierarchies that do not reflect the modern world of work.
LinkedIn has the ability — and obligation — to correct this.
Algorithms should not confine leadership expression. They should expand it.
Because when leadership language becomes more inclusive, leadership itself becomes more inclusive.
**Related Reading: **
Stereobots: When Chatbots Get Typecast — and How We Can Recode Gender in AI