News

LinkedIn is “finally” downranking AI-generated posts

On May 20, 2026, LinkedIn rolled out a detection system that downranks AI-generated content with no human added value. Announced precision of 94%, with automated comments and third-party tools explicitly targeted.

Maxime Ben Bouaziz
CEO Salesdorado
May 25, 2026, updated November 10, 2026, 6 min read
Linkedin déclasse post IA

LinkedIn has just announced a detection system that penalizes the distribution of content deemed “AI slop”, those AI-generated posts with no human added value. Laura Lorenzetti, VP and Executive Editor of LinkedIn Global Editorial, detailed on May 20, 2026 an algorithmic downranking mechanism that also hits automated comments. The system shows 94% precision in testing. Here is what it changes for your sales and marketing teams.

Key takeaways

  • LinkedIn does not delete AI content, but blocks its distribution beyond the first circle of connections
  • The system targets three areas: generic posts, automated comments, mass content creation tools
  • Announced precision of 94%, but no public data on false positives
  • Content creation jumped 14% year over year, the main driver behind the decision
  • Third-party automation tools (Taplio, Podawaa, Waalaxy) are explicitly targeted
  • The verified members filter extends to comments and the feed
  • Full rollout will take several months
SalesdoradoSimulate your LinkedIn prospecting tool selection
Free tool, 2 minutes
Simulate your LinkedIn prospecting tool selection

Answer 11 targeted questions and find the tools that actually fit your situation.

12
tools analyzed
11
targeted questions
€0
no sign-up
Methodology published by the Salesdorado editorial team

What exactly LinkedIn is targeting

The scope of the measure goes beyond posts alone. Lorenzetti identifies three categories penalized by the algorithm. The first covers AI-generated posts that “sound polished on the surface but lack a unique perspective or substance”. The second targets comments produced by automation tools “with little or no human input”. The third includes replies that simply rephrase the original post without adding anything.

According to TechRadar, LinkedIn also targets specific syntactic markers: em dashes and phrasings like “it’s not X, it’s Y” are flagged as signs of LLM output. Shelly Palmer points out the paradox: these patterns existed among human writers long before ChatGPT, and slop generators will adapt their output as soon as the criteria become known.

How the penalty works

No deletion is planned. Detected content stays visible to direct connections and followers, but drops out of algorithmic recommendations. In practice, a flagged post will not travel beyond your immediate network. For a B2B social selling strategy that relies on viral reach beyond the first circle, the impact is direct.

Why is LinkedIn acting now?

The economic motivation is explicit in Lorenzetti’s statements to Entrepreneur. Content creation on the platform is growing 14% year over year, growth directly attributed to the spread of AI tools. If the feed fills up with interchangeable content, engagement drops, advertisers pull back, and the premium model (Sales Navigator, Recruiter, Premium Business) loses its perceived value.

The timing is also competitive. X has damaged the reliability of its verification, Meta and TikTok have tightened AI content labeling, YouTube has demonetized mass-production channels. LinkedIn is positioning itself as the last professional platform where the human voice stays algorithmically protected.

The “AI solving AI” approach

The system relies on supervised training by in-house editorial teams. Human editors and content managers annotate thousands of posts, sorting them into generic and original, based on detailed definitions of low-value and high-value content. The algorithm then learns to reproduce these judgments at scale. The method raises an operational question for brands: who defines what is “original”? The criteria remain opaque, and Lorenzetti herself acknowledges to Entrepreneur that the feed has become more competitive, and that the timing with the rise of AI is obvious.

The concrete implications to factor in

The operational implications are concrete for sales and marketing leaders who have built a LinkedIn presence over the past two years.

  • For employee advocacy programs: The practice of generating posts with ChatGPT and having sales teams publish them becomes risky. If the algorithm detects an unedited AI signature, the post stays published but does not leave the salesperson’s network. The investment in training and tooling loses its leverage. The value goes back to the salesperson who takes the time to write a personal perspective, even if it is stylistically imperfect.
  • For users of automation tools: Taplio, Podawaa, Waalaxy, AuthoredUp: the whole ecosystem is targeted without being named. Automated comments are especially exposed, since the system explicitly targets replies that rephrase the post without adding anything. Presence strategies based on strings of generic comments to gain visibility are dead in the short term.
  • For executives and ghostwriting: Ghostwriting is still allowed as long as it carries a real perspective. A post written by a writer based on an interview with the executive, with its own angle, gets through. A post generated from a generic prompt on a news topic does not. The line shifts to one question: is there lived expertise behind the text?

The stated limits

Several gray areas deserve the attention of advertisers and professional content creators.

The false positive rate is still unknown

The 94% precision figure is put forward with no methodological detail. The Next Web notes that LinkedIn has not shared any data on false positives, and how often legitimate posts will be wrongly flagged as slop remains unknown. For a B2B creator who uses AI as a writing assistant while keeping editorial control, the risk of being wrongly downranked is real.

The rollout will be gradual

Fast Company reports that the rollout will be slow and that it could take several months before all users see less slop in their feed. The effects on distribution will therefore be neither immediate nor uniform across regions and industries.

The Microsoft paradox

LinkedIn belongs to Microsoft, which is actively pushing Copilot, including for writing LinkedIn posts. The Media Copilot notes that LinkedIn itself offers a suite of generative AI tools, including a highly visible “rewrite with AI” button in its post editor, which makes the timing of the announcement awkward. The line between “acceptable AI” and “AI slop” will be political as much as technical.

How to adapt your LinkedIn strategy?

A few practical principles for editorial and sales teams that want to stay visible under the new algorithmic regime.

Keep AI as an assistant, not as the writer. Using ChatGPT to structure an idea or rework a draft is not a problem. Asking it to generate an entire post from a topic is.

Work on signals of lived expertise. An internal figure, specific customer feedback, an operational anecdote, a strong opinion: anything that cannot come out of a generic prompt strengthens the human signal for the algorithm.

Audit your automation tools. If your teams use a third-party tool to churn out comments or posts, the return on investment is going to deteriorate quickly. Reallocating the budget to in-house editorial production becomes an option worth considering.

Get your executives’ accounts verified. The “verified members” filter extended to comments gives a visibility edge to authenticated profiles. For strategic profiles (CEO, VP of sales, directors), verification becomes a distribution asset.

Summary table of the changes

Here is an overview of the main types of content and profiles affected by LinkedIn’s new policy, with how the algorithm treats them before and after the detection system is rolled out.

Content type Status before Status after
Generic AI post with no angle Normal distribution Limited to the first circle
AI post with a human perspective Normal distribution Normal distribution maintained
Automated comment Normal distribution Algorithmic penalty
Comment rephrasing the post Normal distribution Algorithmic penalty
Fake AI profiles Reactive moderation Stronger detection
Verified profiles Limited advantage Filter extended to the feed and comments

What to watch in the coming months

Three indicators will matter to assess the real reach of the measure. First, the initial statements from third-party tool vendors (Taplio, Waalaxy,…) on how they are adapting their technology. Their ability to get around the detection signals will decide the survival of part of the LinkedIn tools market.

Second, engagement data published by LinkedIn and by third-party aggregators such as Favikon or Shield. A measurable drop in the average reach of accounts using heavy automation will be the first sign that the system works.

Third, feedback from B2B creators on false positives. If recognized editorial voices start seeing their reach fall with no explanation, the measure could be revised. One to watch.

★ The conversation, 0 comments

What the community says.

Basic Markdown, moderated within 24 h
Loading…

    Read next