Only 30% of LinkedIn Members Are Active Users. Your Prospect List Does Not Know That.

The LinkedIn number that never makes the press release

LinkedIn tells the market it has 1.3 billion members. On 1 September 2026, buried in a regulatory filing, it told the European Union something considerably more useful: how many of those people actually open the app.

The answer, extrapolated across regions, is roughly a third of them. LinkedIn active users number somewhere around 433 million globally, against a headline member base of 1.3 billion. That gap is not a rounding error or a vanity-metric quibble. It is the single most under-priced fact in B2B prospecting right now, because every list you build, every Sales Navigator search you save, and every "total addressable market" slide you have ever presented is counted in members, not users.

If you run outbound on LinkedIn, this is the number that explains a lot of your Q3.

What the disclosure actually says

LinkedIn's latest EU Digital Services Act disclosure, reported by Social Media Today on 1 September, contains two findings worth a sales leader's attention.

The first is scale. LinkedIn added 1.4 million active EU users in the first half of 2026 versus the second half of 2025, reaching 56.5 million active users in the EU. Set against LinkedIn's reported EU membership, that works out at roughly 30% of members being active in the app. The DSA figure is the only real active-user disclosure LinkedIn makes anywhere — everything else it publishes is member counts, which are cumulative and never decline.

The ratio is also remarkably stable. Social Media Today's reporting on the March 2025 DSA filing put the active share at 28%, and the October 2024 filing put it at 28% as well. So while LinkedIn's member number climbs every quarter and CEO Daniel Shapero reports double-digit member growth and content consumption up 10% year over year, the proportion of members who are genuinely present has barely moved in two years. Engagement is rising among the people already there. The active base itself is growing slowly.

The second finding is enforcement. LinkedIn reported a 46% rise in detected instances of inauthentic activity in H1 2026 versus H2 2025 — automated posting, engagement pods, AI-generated comments and fake profiles. That is a big jump in a six-month window, and it lands on the same platform where your team is running sequences.

Read those two together and the strategic picture is uncomfortable: the reachable audience is roughly a third of what your tooling implies, and the enforcement surface around automated outreach is expanding fast.

Members are not users, and your prospecting list cannot tell the difference

Here is the practical problem. LinkedIn search — including Sales Navigator — indexes profiles. A profile exists whether its owner logged in this morning or in 2019. Filter by title, headcount, geography and industry and you get a count of profiles that match, which your team then treats as a count of people you can reach.

Those are different numbers, and the difference is roughly 3x.

The arithmetic on a 10,000-person list

Run a Sales Navigator search for "Head of Revenue Operations, 200-1,000 employees, US, SaaS" and suppose it returns 10,000 results. Your SDR manager sees ten thousand prospects and a quarter's worth of work. What is actually in there?

Now layer on LinkedIn's sending limits. Most accounts sit at roughly 100 connection invitations per week. Working a 10,000-person list at that rate takes about two years — and on the numbers above, roughly seven out of every ten of those invitations go to somebody who is not meaningfully present on the platform. You are not being rejected. You are talking to an empty room, and paying for the privilege in invitation quota, in sender reputation and in your team's morale.

That is also why "just add more accounts" has been such a poor fix. Adding seats multiplies the invitation quota you can spend, but it does nothing about the composition of the list you are spending it on. Three times the sending capacity against a list that is 70% dormant produces three times the dormant sends.

Why titles are the worst filter you have

Title-and-firmographic filtering optimises for the wrong variable. It selects for who someone is on paper, which is exactly the attribute that persists on a profile long after the person stops using the platform, changes role, or delegates the category entirely. It tells you nothing about presence.

Worse, everyone else is running the same filters. If your ICP is "VP Sales at Series B SaaS in North America," so is every competing vendor's, which means the small active slice of that segment is absorbing outreach from dozens of teams while the dormant majority absorbs none. The active people are over-messaged and the inactive people are unreachable. Both halves of the list are working against you, in opposite directions.

Part of the reply-rate collapse is a denominator problem

The prevailing story in B2B sales for the last eighteen months has been that LinkedIn outreach is dying. Expandi's analysis of 13.2 million data points found connection-request reply rates falling from 3.5% in May 2025 to 2.2% in April 2026. Email is no better: the platform-wide reply average sits around 3.43% according to the benchmarks compiled in ORRJO's State of B2B Outbound 2026.

Those numbers are real. But look at how they are computed. Reply rate is replies divided by messages sent, and messages sent includes every send into a dormant account. If roughly 70% of a typical scraped list is inactive, then a 2.2% headline reply rate against all sends corresponds to something in the region of 7% against sends that actually reached a live user.

That reframing matters more than it looks, for three reasons:

• It changes the diagnosis. If your reply rate is falling because the channel is saturated, the answer is better copy. If it is falling because your denominator is filling with dormant profiles, better copy will not move it at all — you will write beautiful messages into inboxes nobody opens.
• It changes the economics. Cost per meeting is calculated on total sends. Strip the dormant 70% out of the denominator and the cost per reachable contact triples, which is the real number your board should see.
• It changes what "scale" means. Volume tooling scales the denominator. It does not scale the reachable population, which is fixed by how many people in your ICP actually open LinkedIn.

The caveat that keeps this honest

Two things temper the arithmetic, and it is worth being straight about them.

First, the ~30% active share is an all-member average, and B2B decision-makers in tech-adjacent roles are almost certainly more active than the platform mean. Nobody has published a segment-level active rate, so the true figure for your ICP is higher than 30% — probably meaningfully higher, but unknown. Treat 30% as a floor and a warning, not a precise coefficient for your specific segment.

Second, "active" in the DSA sense is a monthly measure, not a measure of whether someone reads messages. Plenty of monthly-active users log in to check a notification and never open their inbox. The functional reachable population for outreach is smaller than the active-user count, not larger. The direction of the error is consistent: every step from member to active user to someone who reads messages to someone who replies shrinks the pool, and every one of those steps happens after your list tool has already given you its confident count.

The other half of the report: enforcement is up 46%

While the addressable side is smaller than teams believe, the risk side is expanding. LinkedIn's 46% rise in detected inauthentic activity is not an abstract trust-and-safety statistic — it describes the detection systems your outbound tooling now has to survive.

What LinkedIn is counting

"Inauthentic activity" in LinkedIn's framing spans several things that are adjacent to, or indistinguishable from, standard automated outbound:

• Fake and duplicate profiles
• Engagement pods and coordinated fake engagement, which LinkedIn publicly committed to acting against in late 2025
• External apps offering automated posting and messaging services
• AI-generated engagement that adds nothing to the feed — the category behind the "seems like AI slop" report button, which passed a million uses in three weeks

The scale of the underlying problem explains the urgency. An analysis by AI detection startup Pangram Labs, across 57,000 public LinkedIn posts, found that roughly 30% of all comments posted between April and June 2026 were entirely AI-generated. LinkedIn's own Q2 numbers showed time spent in post comments up 18% year over year — a rise that looks a lot less impressive once you know a third of the comments were written by machines.

Why volume tooling gets flagged first

Detection systems do not evaluate your intent. They evaluate patterns: session behaviour that does not look human, message templates repeated across accounts, engagement that arrives in coordinated bursts, connection requests fired at a machine-steady cadence. A team sending 100 near-identical invitations a week through a browser-automation tool generates exactly the signature LinkedIn is investing in catching, regardless of how legitimate the underlying business is.

The strategic conclusion writes itself. Any outbound motion whose only lever is volume is now fighting two trends at once: a reachable population roughly a third the size of its list, and an enforcement apparatus that got 46% better at spotting mechanical behaviour in six months. You cannot out-send either of them.

The fix: proof-of-life targeting

If the constraint is presence rather than fit, then the highest-leverage change available to most teams is not a new template or a new tool. It is switching the primary filter from who someone is to what someone recently did.

Every observable action on LinkedIn is a proof of life. Someone who commented on a post yesterday is, definitionally, in the active 30%. Someone who viewed your profile this week has opened the app and is looking at people. Someone whose company posted a job spec on Monday has a team spending money on a problem right now. None of these require you to guess whether the person exists — the action is the evidence.

The signal hierarchy, ranked by what it proves

Not all signals are equal. Ranked by how much they prove about both presence and intent:

• Profile views of you or your teammates. The strongest single signal available. It proves activity, and it proves attention pointed at your category, this week.
• Engagement on a relevant post — theirs or someone else's. Proves presence within days, and the post topic tells you what they care about. Engagers on a competitor's post are the highest-value version of this.
• Public complaints and questions about the problem you solve. A Reddit thread asking for tool recommendations, an X post venting about a workflow, a LinkedIn post describing the pain. Proves the person is active somewhere and has stated the need in their own words.
• Hiring signals. A job posting for a role adjacent to your product proves budget and organisational intent even when the individual buyer is quiet. Weaker on presence, strong on timing.
• Funding, leadership changes and job changes. Classic trigger events. They prove that something changed at the account, not that any particular person is reachable, so pair them with an activity signal before you send.
• Firmographic fit alone. Necessary, never sufficient. This is the filter that produced the 10,000-person list.

The practical rule: fit qualifies an account, activity qualifies a contact. A prospect should clear both bars before they enter a sequence. This is precisely the model Updately is built around — capturing warm signals like profile views, post engagers, competitor mentions and hiring activity, scoring them against your ICP, researching each prospect properly, and only then sending inside LinkedIn's own limits. The point is not to send more. It is to make sure that what you do send lands on someone who is actually there.

Rebuilding a list you already have

You do not need to start over. Take your existing lists and apply an activity pass:

1. Segment by observable recency. Anyone who has posted, commented, or changed roles in the last 90 days goes into an active tier. Everyone else goes into a hold tier — not deleted, just not sequenced.
2. Sequence the active tier first, at low volume and high specificity. These are the people who can reply. Spend your personalisation budget here.
3. Convert the hold tier into a listening problem. Instead of messaging them, watch for a signal. When a held contact engages with something relevant, they promote themselves into the active tier and get sequenced within 24 to 48 hours.
4. Add a second channel for the accounts that matter. If an account clears the fit bar but nobody there is active on LinkedIn, that is a routing decision — email, phone, or a warm path through a mutual connection — not a reason to keep spending invitations.
5. Retire the vanity denominator. Stop reporting list size. Report reachable contacts.

The signal-led version of this is measurably better where it has been tested. ORRJO's benchmark work found a signal-based cohort booking 3.4x more meetings with 60% fewer touchpoints, at an 11.2% reply rate against 2.1% for a volume-led group. That spread is not explained by copy quality. It is explained by who was on the receiving end.

What this does to sequence design

A smaller, verified-live list changes the shape of a sequence. With 300 reachable contacts rather than 10,000 nominal ones, the constraint stops being capacity and starts being quality, which means:

• Front-load the effort. Reply rates decay sharply after the second touch across every dataset worth reading, so put your research into the opener rather than into follow-up six.
• Cut sequence length. Four to six touches over three to four weeks is plenty when the list is qualified. Ten-touch sequences exist to compensate for bad targeting.
• Reference the signal explicitly. If someone commented on a post about pipeline forecasting, say so. It is the cheapest proof that a human read something, and it is the one thing a dormant-list blast structurally cannot do.
• Keep a human in the approval loop for your top accounts. It costs minutes on a 300-person list and is impossible on a 10,000-person one.

Metrics to change this week

Three reporting changes follow directly from the DSA numbers, and all three can be made before your next pipeline review.

The second row is the one that will start arguments, and it is the one that matters most. A team reporting 2% replies against a bloated denominator looks like it has a messaging problem. The same team reporting 7% against a live denominator looks like it has a sourcing problem, which is both true and fixable.

Takeaways

• LinkedIn's H1 2026 EU DSA disclosure implies only around 30% of LinkedIn members are active users — roughly 433 million globally against 1.3 billion members. That ratio has been flat at 28-30% since late 2024.
• Sales Navigator and every scraping tool count members, not users. A list of 10,000 matching profiles likely contains around 3,000 reachable people, and far fewer who read messages.
• A meaningful share of the "LinkedIn reply rates are collapsing" narrative is a denominator problem. Reply rate over live contacts is roughly 3x the headline figure — which reframes the fix from better copy to better sourcing.
• Detected inauthentic activity on LinkedIn rose 46% in H1 2026, and roughly 30% of comments in Q2 were entirely AI-generated. Volume-first automation is now fighting a smaller reachable audience and a rapidly improving detection system simultaneously.
• The response is proof-of-life targeting: use firmographics to qualify the account and observable recent activity to qualify the contact. Profile views, post engagement, public complaints and hiring signals all prove presence in a way a job title never can.
• Change the reporting first. If your dashboard still shows list size and gross reply rate, your team is being measured on a number that is three times larger than reality.

The uncomfortable truth in this disclosure is that most outbound teams have been optimising the wrong half of the funnel. Copy, cadence and tooling have all improved enormously in two years. The list has not. If roughly seven in ten of the people on it were never going to see the message, no amount of craft further down the sequence was ever going to matter.