LinkedIn by the numbers — the platform creators underrate

You see someone's LinkedIn post get 50,000 views. You post the same day. Yours gets 400. You look at what they wrote. You look at what you wrote. You try to spot the difference.

So you go looking for advice. Most of it is about the algorithm. What it wants. How to please it. Almost none of it quotes what LinkedIn actually publishes about how any of this works. And almost none of it starts with the shape of the platform itself.

Here is what the numbers say. LinkedIn has more than 1.3 billion members. Fewer than one in four use it monthly, which leaves a monthly audience roughly a tenth the size of Instagram's three billion. One percent of those monthly users post content weekly. That one percent generates 9 billion impressions every week. The platform pulls in nearly $20 billion a year. And in March 2026, LinkedIn replaced the whole ranking system with one built on large language models and graphics processors.

What follows is drawn from Microsoft's filings and LinkedIn's own reported metrics. It covers the size of the audience, how much money the platform makes, where that money comes from, and what LinkedIn has published about the new feed system. Every source is linked, dated and quoted. One section describes a limit we could not work around. We start with who is actually there.

The gap between members and monthly users is huge

LinkedIn counts two kinds of number. Members are accounts. That is everyone who signed up and has not deleted their account. Monthly active users are the people who came back and used it in the last month. The gap between them is the whole story.

Start with members. Microsoft's annual report for fiscal year 2025 gives the count. The fiscal year ends 30 June 2025. The report says LinkedIn "is now home to 1.2 billion members". LinkedIn's own Q4 FY2026 earnings note, published in July 2026, says "global membership grew at a double-digit rate for the fifth consecutive year". It gives no number. Microsoft's FY2026 ends 30 June 2026, so that Q4 figure is for the quarter ending that date. A third-party estimate puts membership at 1.3 billion by late 2025. We use 1.2 billion from the Microsoft report because it is the only official count we read.

Monthly active users are different. LinkedIn does not publish that number. Third-party sources put it at around 310 million. That figure appears across multiple statistics pages in 2026, all citing an external research firm. We could not read the original source those pages cite. So we report 310 million as an estimate, not as a LinkedIn-published figure.

If that estimate is right, it means one in four members uses LinkedIn monthly. Three in four signed up and do not come back. Or come back so rarely they do not count as active in a month.

Daily active users sit even lower. The same third-party estimate puts daily users at roughly 134 million. That would be about 43 percent of monthly users. And about 11 percent of all members.

Every other platform's funnel looks like this too. Signing up is easy. Coming back is not. What matters here is the width of the gap. A pool of 1.2 billion members sounds vast. A pool of 310 million monthly users is still big. But it is not the same pool. Your post is competing for attention inside the smaller one.

Only one percent post content weekly

Here is the number that changes how you read everything else. Around one percent of LinkedIn's monthly active users post content weekly. Those users generate 9 billion impressions per week.

That claim appears in multiple statistics round-ups published in 2025 and 2026. All of them trace it back to session data LinkedIn published in 2019. We went looking for the original LinkedIn source. We could not find it. The 2019 LinkedIn engineering and company blogs we checked do not state this figure. One third-party marketing page from 2019 says "over nine billion impressions each week" but does not link to a LinkedIn document. We do not cite marketing blogs as sources.

So we report the claim as widely repeated, with its origin described as 2019 LinkedIn session data, but we could not verify it from a primary source. That makes it the weakest number in this post.

If it is true, the math is stark. One percent of 310 million is 3.1 million people. Three million people posting. More than 300 million reading. Your post is not fighting for space in a crowd of a billion creators. It is fighting for space in a crowd of three million.

And nine billion impressions divided by three million creators is 3,000 impressions per creator per week. That is an average. The distribution will not be flat. Some will get far more. Most will get far less.

LinkedIn made $19.8 billion in fiscal 2026

Revenue is the one number both Microsoft and LinkedIn report clearly. LinkedIn's Q4 FY2026 earnings note says full-year revenue hit $19.8 billion. That is for the fiscal year ending 30 June 2026. Revenue grew 12 percent year over year in Q4.

Microsoft breaks LinkedIn's revenue into four lines. Talent Solutions is recruiting tools for employers. Marketing Solutions is ads. Sales Solutions is tools for salespeople. Premium Subscriptions are paid member accounts. Microsoft's fiscal 2025 annual report says LinkedIn revenue grew 9 percent that year, "driven by growth across all lines of business". The FY2026 Q3 earnings report says LinkedIn revenue increased by $521 million, or 12 percent, "with growth across all lines of business".

Microsoft does not publish how much each line contributes. It only says all four grew. So we know LinkedIn makes money from recruiting, ads, paid tools for sales teams, and paid accounts. We do not know which one is biggest.

One number breaks out. LinkedIn Marketing Solutions revenue grew 16 percent year over year in Q4 FY2026. That is LinkedIn's ad business. It recorded "seventh consecutive quarter of double-digit growth". So ads are growing faster than the platform's overall revenue.

The other detail worth holding: Microsoft's FY2025 annual report says LinkedIn revenue "is mainly affected by demand from enterprises and professionals for subscriptions to Talent Solutions, Sales Solutions, and Premium Subscriptions offerings, as well as member engagement and the quality of the sponsored content delivered to those members to drive Marketing Solutions". That sentence names enterprises and professionals. It does not name small businesses or creators as the revenue driver. LinkedIn sells to companies and to individuals who pay for premium features. Creators are the content. Not the customer.

Engagement is up but the platform fights fake interaction

LinkedIn reported three engagement metrics in its Q4 FY2026 earnings note. Time spent on content grew 10 percent year over year. Knowledge-oriented posts increased 24 percent. Time spent reading comments rose 18 percent. All three point the same way. People are reading more and staying longer.

The same note from July 2024 said members engage with 1.5 million pieces of content every minute. Video was the fastest growing format. Uploads were up 34 percent year over year at that point.

But LinkedIn also said in March 2026 that it is cracking down. The company published a news post titled "How LinkedIn Is Improving the Feed to Show More Relevant, Authentic Professional Content". It lists four changes. The first is smarter ranking using large language models. The second is fighting fake engagement. LinkedIn named three things it is removing: "comment automation, engagement pods, and unauthorized third-party tools". Engagement pods are groups that agree to like and comment on each other's posts to game the system. The goal, LinkedIn says, is making sure users interact with "a real person and a real point of view".

The third change is reducing low-quality content. LinkedIn called out "repetitive, low-substance posts and engagement bait" like "comment to agree" prompts. The fourth is better onboarding for new members. LinkedIn is testing an interest picker during sign-up so people can tell the system what they want to see.

That is LinkedIn saying the platform had a fake engagement problem big enough to name publicly. And that it is changing the ranking system to fix it. If you bought engagement, or joined a pod, or used automation, the March 2026 changes are aimed at you.

The new feed runs on language models and GPUs

LinkedIn rebuilt the entire feed system in March 2026. The company published a technical post on its engineering blog that month. The post is the most detailed thing LinkedIn has written about how the feed works.

The old system used multiple separate retrieval methods. Each had its own infrastructure. The methods included a chronological feed of your network's posts, trending content in your region, recommendations based on what similar members liked, industry-specific trends, and several embedding-based systems. Embedding means writing a short description of you and a short description of a post as numbers, then comparing the numbers to see if they match.

The new system replaced all of that with two parts. A dual-encoder model powered by a large language model pulls together the posts worth considering. Then a sequential transformer model called Generative Recommender ranks them. Both run on GPU clusters. GPUs are graphics processors. They are faster than regular processors for this kind of math.

The dual encoder works like this. It takes a description of you and a description of a post. Both descriptions are written by the same language model. The model turns each description into a set of numbers called an embedding. Then the system compares the two embeddings using a measure called cosine similarity. Posts with embeddings close to yours rank higher.

One technical detail matters. Raw engagement counts did not work. A post with 500 likes and a post with 50 likes would just get scored by those numbers. The correlation with relevance was nearly zero. LinkedIn fixed it by converting counts into percentile buckets. A post in the 71st percentile for views gets tagged with a token that says so. That change increased correlation 30 times. And it improved recall at 10 by 15 percent. Recall at 10 means how many of the top 10 results were actually relevant.

The ranking model reads your history. LinkedIn says it processes "more than a thousand of your historical interactions". It treats your feed as a sequence, not as a set of independent events. So it knows you liked a post about hiring, then read a post about remote work, then skipped a post about sales. That order matters. The model uses transformer architecture with causal attention. Causal means each prediction can only look backward, not forward.

The system runs three continuous pipelines to keep embeddings fresh. Prompt generation updates descriptions of members and posts within minutes as activity happens. Embedding generation runs the language model on GPU clusters to turn those prompts into numbers. GPU-accelerated indexing builds the search index that matches you to posts. The system retrieves results in under 50 milliseconds while searching across millions of indexed posts.

LinkedIn says the new system eliminated engineering complexity, enabled more coherent candidate sets for ranking, and made state-of-the-art models economically viable for all 1.3 billion members. It does not say the new system shows more posts to more people. It says it shows more relevant posts.

What this means for someone posting

Combine the numbers. You are one of 3.1 million people posting weekly into a feed read by 310 million monthly users. Your post is being ranked by a system that reads your post semantically, compares it to a reader's history of more than a thousand interactions, and decides in milliseconds whether it is relevant. The system was rebuilt from scratch seven months ago. And LinkedIn is actively fighting fake engagement.

The advice you see everywhere is about tricks. Post at this time. Use this format. Add this emoji. That advice assumes the system is simple and can be gamed. The system is not simple. It is a pair of transformer models running on GPU clusters, trained on billions of interactions, reading the semantic content of your post and the behavioral history of every reader. You cannot trick it with an emoji.

What you can do is narrower. Write something a specific kind of professional wants to read. Make it knowledge-oriented, which is the term LinkedIn used when it said those posts grew 24 percent. Do not beg for engagement. LinkedIn named that as something it is filtering out. Do not join a pod. LinkedIn named that too. And do not expect reach like Instagram. The creator pool is smaller. The audience is smaller. The use case is different. People come to LinkedIn to hire, get hired, learn, or sell. Not to be entertained.

The platform underrates creators in one sense and overrates them in another. It underrates them because LinkedIn makes money from enterprises, not from creators. Creators are the content, not the customer. It overrates them because only one percent post, so the bar to join that one percent is lower than on any other platform. The question is not whether you can go viral. The question is whether your 3,000 impressions per week reach the people who matter to you.

What we could not verify

Three things, stated so you know where this post stops.

The original 2019 LinkedIn source for the "9 billion impressions" and "1% of users post weekly" claim. Every statistics page cites it. We could not find the original LinkedIn document. We checked LinkedIn's 2019 engineering blogs, company news posts, and marketing solution blogs. None of them state this figure in the pages we read. One third-party marketing blog from 2019 mentions "over nine billion impressions each week" but does not link to a source. So we report the claim as widely repeated but unverified from a primary source. If LinkedIn published it, we could not find where.

The exact split of revenue across the four business lines. Microsoft reports total LinkedIn revenue and says all four lines are growing. It does not publish what percentage comes from Talent Solutions versus Marketing Solutions versus Sales Solutions versus Premium Subscriptions. We only know Marketing Solutions grew 16 percent in Q4 FY2026 because LinkedIn broke that one out in an earnings note.

The exact daily active user count. We report 134 million as a third-party estimate. LinkedIn does not publish this number. The estimate comes from an assumption, not a measurement. The firm that publishes it derives daily users by applying a 16.2 percent ratio. We could not verify that ratio from a LinkedIn source. So we report it as an estimate and say clearly it is not measured.

How we did this

Scope: this post reads published documents. No data collection. No reverse engineering. No experiment of our own.

How we read them: each source was opened directly on 18 August 2026. Every quoted sentence was copied as published. Six sources are text generated by code in the browser, so they were read in a browser. The Microsoft annual report and earnings releases are delivered as finished pages. The LinkedIn engineering blog post is a finished page. The LinkedIn news posts are finished pages. No source was paraphrased from a summary. Where a claim is repeated across multiple third-party pages but we could not find the original source, we say so.

Limits: most sources are first-party statements. Microsoft's filings and LinkedIn's own news posts and engineering blog. A first-party statement is what the company chose to publish. Not an audit. Not an independent measurement. Monthly active users and daily active users are third-party estimates, not LinkedIn-published figures, and we state that each time we use them. The "9 billion impressions" and "1% post weekly" claims are widely repeated but we could not verify them from a primary LinkedIn source. We report them with that caveat rather than pretending we read the original.

Sources

1. Microsoft Annual Report, Fiscal Year 2025 (year ending 30 June 2025). Retrieved 18 August 2026 from https://www.microsoft.com/investor/reports/ar25/index.html. Source of the 1.2 billion members figure, the four business lines, the 9% revenue growth in FY2025, and the statement that revenue is "mainly affected by demand from enterprises and professionals".

1. LinkedIn Q4 FY2026 Earnings and Business Highlights, published July 2026. Retrieved 18 August 2026 from https://news.linkedin.com/2026/q4-earnings-and-business-highlights. Source of $19.8 billion full-year revenue, 12% year-over-year revenue growth in Q4, double-digit membership growth for fifth consecutive year, 10% growth in time spent on content, 24% increase in knowledge-oriented posts, 18% rise in time spent reading comments, Marketing Solutions 16% growth, and seventh consecutive quarter of double-digit Marketing Solutions growth.

1. Microsoft Investor Relations, FY-2026-Q3, Productivity and Business Processes Performance. Retrieved 18 August 2026 from https://www.microsoft.com/en-us/investor/earnings/fy-2026-q3/productivity-and-business-processes-performance. Source of the $521 million revenue increase and 12% growth figure for Q3 FY2026, with growth across all business lines.

1. LinkedIn Business Highlights from Microsoft's Q4 FY24 Earnings, published 30 July 2024. Retrieved 18 August 2026 from https://news.linkedin.com/2024/July/LinkedIn_Business_Highlights_from_FY24_Q4_Earnings. Source of "members engage with 1.5 million pieces of content every minute", video as fastest growing format, and 34% year-over-year increase in video uploads. Also source of 51% increase in Premium sign-ups in FY24.

1. LinkedIn Engineering Blog, "Engineering the next generation of LinkedIn's Feed", published 12 March 2026, authored by Hristo Danchev. Retrieved 18 August 2026 from https://www.linkedin.com/blog/engineering/feed/engineering-the-next-generation-of-linkedins-feed. Source of all technical details about the dual-encoder LLM system, Generative Recommender sequential transformer, GPU infrastructure, the raw engagement count correlation problem (-0.004) and the percentile bucket solution (30× improvement), recall@10 improvement (15%), the "more than a thousand of your historical interactions" statement, transformer architecture with causal attention, sub-50ms retrieval latency, and the three continuous pipelines (prompt generation, embedding generation, GPU-accelerated indexing). Also source of the old system description listing chronological feeds, trending content, collaborative filtering, industry-specific trends, and multiple embedding-based retrieval systems.

1. LinkedIn News, "How LinkedIn Is Improving the Feed to Show More Relevant, Authentic Professional Content", published March 2026. Retrieved 18 August 2026 from https://news.linkedin.com/2026/ImprovingTheFeed. Source of the four changes: smarter content ranking using Generative Recommenders and LLMs, fighting fake engagement (comment automation, engagement pods, unauthorized third-party tools), reducing low-quality content (repetitive posts, engagement bait, "comment to agree" prompts), and better onboarding with interest picker. Also source of "a real person and a real point of view" quote and "the most trusted place for professional, authentic conversations" framing.

1. Third-party estimate of 310 million monthly active users and 134 million daily active users, appearing across multiple statistics aggregation pages in 2026 (demandsage.com, searchscope.com.au, martal.ca, leadfeeder.com, posteverywhere.ai, and others), all citing Business of Apps as the source. We could not access the original Business of Apps source document. The daily figure is described on at least one of these pages as "derived from a 16.2% assumption rather than measured". We report both figures as third-party estimates, not LinkedIn-published numbers. Retrieved 18 August 2026.

1. "Around 1% of LinkedIn's monthly active users post content weekly, generating 9 billion impressions per week" claim, appearing in contentin.io/blog/linkedin-content-statistics (updated July 2026), attributed to "the most recent session data LinkedIn published, in 2019". Also appearing in multiple Forbes articles by Jodie Cook (April 2025, May 2026), insidea.com, linkmate.io, authoredup.com, autoposting.ai, and others. We searched LinkedIn's 2019 engineering blogs, company news, and marketing solution blogs and could not locate the original source. One 2019 third-party marketing page (bluetext.com) says "over nine billion Impressions each week" with no link. We report the claim as widely repeated but could not verify it from a primary LinkedIn source. Retrieved 18 August 2026.

1. Third-party estimate of 1.3 billion members by late 2025, appearing across multiple statistics pages. Microsoft's official FY2025 annual report (source 1 above) gives 1.2 billion as of 30 June 2025. We use the Microsoft figure as the verified count and note the 1.3 billion estimate as a later third-party figure we could not verify from LinkedIn or Microsoft directly.

Changelog

• 2026-08-18: First version drafted.