LinkedIn buried a number in a disclosure filing, and Social Media Today dug it out: detected inauthentic activity on the platform rose 46% in the first half of 2026 compared to the second half of 2025. That’s the headline stat. It’s also the least interesting part of what LinkedIn’s been building over the past five weeks.

Since July 30, LinkedIn has launched a public report button, watched it get clicked over a million times, started quietly warning individual authors in their own analytics that other members think their post “seems like AI,” and killed its own AI writing assistant outright. Four moves, one direction. If you’ve been leaning on AI to draft LinkedIn posts — not touch them up, draft them — the platform just made that a lot riskier without telling you exactly where the line is.

Quick Verdict

	What You Need to Know
The headline number	Detected inauthentic activity up 46% in H1 2026 vs. H2 2025, per Social Media Today
Public report button	”Seems like AI slop,” launched July 30, 1M+ clicks in the first two weeks
Reach hit on flagged posts	~40% fewer views on content LinkedIn’s classifiers tag as slop
New this cycle	A private analytics flag telling authors, in their own dashboard only, that members think a specific post “seemed like AI”
What got killed	The “enhance your post” AI writing tool — replaced with a proofreading-only feature that fixes grammar but won’t draft or rewrite for you
Report threshold disclosed	None. LinkedIn hasn’t said how many flags trigger the private warning
Public penalty for the flag	None disclosed — it’s framed as feedback, not a strike, but nobody’s said what happens if you get flagged repeatedly

Four Moves in Five Weeks

Easy to treat this as one announcement. It isn’t. It’s an escalation that happened in stages, and the stages matter because they show LinkedIn moving from “let users flag it” toward “we’ll tell you ourselves, privately, whether you asked or not.”

1. July 30 — the button and the funeral. LinkedIn added a “Seems like AI slop” option to the three-dot menu on every post and comment, and in the same announcement said it was pulling its own “enhance your post” feature — the tool that used AI to rewrite your drafts. The replacement proofreads. It corrects grammar and phrasing. It does not generate sentences for you anymore.
2. Late August — the numbers came in ugly for slop posters. The button passed a million clicks inside two weeks, and LinkedIn’s chief product officer Hari Srinivasan told Fortune the company still had “more to do.” Posts its classifiers tagged as slop were pulling in roughly 40% fewer views than comparable content before the button existed.
3. Late August — the private flag started rolling out. LinkedIn began testing a feature that tells authors, inside their own post analytics and nowhere else, when enough members have indicated a post reads as AI-generated or inauthentic. It’s not visible to your audience. It’s not a public label. It’s a note that shows up only for you, after the fact.
4. September 1 — LinkedIn confirmed the trend line. The 46% increase in detected inauthentic activity, disclosed as part of routine EU transparency reporting, put a company-level number behind five weeks of individual product changes.

Read in order, this isn’t a policy tweak. It’s a platform deciding that self-reporting plus a public button wasn’t enough, and adding a private surveillance layer on top of it while simultaneously cutting off the tool that made a lot of the slop in the first place.

The Private Flag Nobody’s Explained

Here’s where this gets uncomfortable if you post on LinkedIn regularly. The private analytics warning is real, it’s live for at least some accounts, and LinkedIn has told reporters almost nothing about how it actually triggers.

We don’t know how many “seems like AI slop” reports it takes before the flag shows up in your dashboard. We don’t know if it’s a raw report count, a ratio against your post’s total views, or something LinkedIn’s classifiers weigh independently of user reports at all. We don’t know whether the flag stacks — does getting flagged three times in a month change anything, or does each post reset to zero? None of that is published. LinkedIn’s own framing, reported by outlets covering the test, is that this is meant as feedback: “We want members to get feedback from real humans on what sounds authentic, not just have an AI detector review it and get it wrong.” That’s a reasonable goal. It’s also a company telling you to trust a system it won’t describe.

Compare this to how Instagram handled its AI-generated profile penalty a few weeks earlier — a real reach cut, but at least a clearly stated trigger (AI-generated person, undisclosed) and a clearly stated consequence (excluded from non-follower recommendations). LinkedIn’s version tells you people are unhappy with a specific post and stops there. No stated penalty. No stated appeal. Just a private note that something about your writing read as fake to enough people that LinkedIn decided to mention it.

What Actually Killed “Enhance Your Post”

The old feature took your draft and rewrote it — punchier hooks, restructured paragraphs, the kind of AI polish that made a two-line thought look like a LinkedIn thought leader post. That’s gone. What replaced it proofreads: it catches typos and awkward phrasing without touching your structure or your voice, per LinkedIn’s own description of the change.

If your LinkedIn workflow was “type a rough idea, hit enhance, post whatever came back,” that workflow doesn’t exist on LinkedIn’s own tools anymore. You can still paste a draft into ChatGPT or Claude and bring it back — LinkedIn hasn’t banned external AI use, and it’s said explicitly that using AI to refine your own ideas isn’t the target. But the platform removing its in-house rewrite tool, in the same breath as launching a public slop-reporting button, isn’t a coincidence. It’s LinkedIn taking itself out of the business of generating the exact kind of content it’s now penalizing.

The Numbers So Far

Three figures, three different things they measure. Worth keeping straight:

• 1 million+ — clicks on the “Seems like AI slop” button in its first two weeks, confirmed by Fortune. This measures user willingness to report, not accuracy.
• ~40% — the drop in views on posts LinkedIn’s classifiers tag as slop, compared to before the button launched. This is the actual reach consequence, and it’s driven by a mix of user reports and LinkedIn’s own detection, not user reports alone.
• 46% — the year-over-half-year rise in detected inauthentic activity overall, from LinkedIn’s EU transparency disclosure. This measures LinkedIn’s detection getting more aggressive, engagement pods and automated posting included — not just AI-written text.

None of these numbers tell you your specific odds of getting flagged. They tell you the system got a lot busier in a short window, and busier systems make more mistakes, not fewer.

How Many Reports Does It Take to Get Flagged?

LinkedIn hasn’t published a threshold. The private analytics warning appears to combine user reports from the “Seems like AI slop” button with LinkedIn’s own content classifiers, but the company hasn’t disclosed a report count, a percentage of viewers, or any other number that triggers the notice. Until LinkedIn says otherwise, treat every post as evaluated individually rather than assuming a fixed number of clicks sets it off.

Who Actually Needs to Worry

Creators who used “enhance your post” to draft, not edit. If your process was rough idea in, polished post out, with LinkedIn’s own AI doing the heavy lifting, you’ve lost that tool and you’re now posting into a system actively hunting for the output it used to produce.

Anyone running a high-volume LinkedIn content calendar built on AI drafting tools. Volume posting — three, four, five posts a week, all AI-assisted, all following the same hook-story-lesson template — is exactly the pattern classifiers are built to catch. We’ve flagged this same shape before: platforms move fast once detection tech catches up to a workaround creators found first.

B2B creators building a LinkedIn-first strategy right now. If you’re chasing a BrandLink-style monetization path or trying to build the kind of consistent presence LinkedIn rewards, getting privately flagged repeatedly is exactly the kind of signal that could work against you later, even without a stated public penalty today.

Anyone who assumes “private” means “harmless.” A flag only you can see is still a flag. LinkedIn hasn’t said the private warning has zero downstream effect on distribution, and given the company’s own classifiers already cut reach by 40% on posts identified as slop, treating the private flag as consequence-free feels optimistic.

Who’s probably fine: creators using AI for research, outlining, or light editing while writing the actual post themselves. LinkedIn has been consistent that refining your own ideas with AI assistance isn’t the target — full generation dressed up as a personal take is.

This is the same split platforms keep landing on. YouTube’s AI-content detection and Spotify’s AI persona badge both draw the same line: tools that assist a real person are fine, tools that fabricate a persona or a voice aren’t. LinkedIn’s version of that line is just blurrier right now, because the trigger for its private flag isn’t published anywhere.

What to Do About It Right Now

1. Stop pasting full AI drafts straight to publish. Use AI for outlining or grammar, same as LinkedIn’s new proofreading tool does, and write the actual sentences yourself. The gap between “AI helped me think” and “AI wrote this” is exactly what’s being policed.
2. Check your analytics dashboard regularly, not just your engagement numbers. The private flag lives there. If you’re not looking, you won’t know you’ve been warned until reach quietly drops on future posts.
3. Vary your structure. Classifiers key on patterns — identical hook formats, identical “here’s what I learned” closers, identical line-break rhythm across every post. Sameness reads as automated even when a human wrote every word.
4. Don’t assume the report button only catches obvious slop. A million clicks in two weeks means plenty of borderline posts got flagged too. If your engagement rate drops without an obvious reason, a slop report is now a real candidate explanation.
5. Reread your last ten posts as a stranger would. If you can’t tell whether a real person or a language model wrote a paragraph, LinkedIn’s classifiers probably can’t either — and they’re erring toward flagging.

Our Take

The report button was the easy, defensible part of this rollout back in July — give users a way to say “this feels fake,” let the platform sort signal from noise. Killing “enhance your post” is defensible too, if a little late; LinkedIn was arguably contributing to its own slop problem with that tool. Neither of those moves is the part we’d push back on.

The private analytics flag is where LinkedIn is asking for trust it hasn’t earned yet. Telling an author, with no explanation of the trigger and no stated consequence, that people think their writing “seems like AI” is a strange kind of transparency — it looks like disclosure, but it discloses nothing about how the underlying judgment got made. We’ve seen this pattern before, on other platforms, in this same stretch of 2026: policies that name the mechanism and skip the specifics. LinkedIn at least deserves credit for testing something less punitive than an outright reach cut. But “here’s a private warning with no rules attached” isn’t actually less confusing than a public penalty — it’s just quieter about it.

If you’re posting on LinkedIn for work, for a business, or for anything you’d call a personal brand, the practical move is the same either way: write the post yourself, use AI the way you’d use a good editor, and check your dashboard before you assume a flat week of engagement is just an off week.

---

Sources: Social Media Today — “LinkedIn increases push against inauthentic activity” (Sept. 1, 2026), TechCrunch — “LinkedIn adds a button to report AI-generated ‘slop’” (July 30, 2026), Fortune — “Over 1 million people clicked LinkedIn’s ‘seems like AI slop’ button” (Aug. 25, 2026). Report thresholds and the private flag’s exact trigger have not been published by LinkedIn as of this post’s publish date.