Daily Marketing News, October 10, 2026: LinkedIn explained how it hides AI slop from the feed
On October 9 LinkedIn explained how it hides templated posts written by neural networks from the feed. The same day X taught its AI agent to track brand mentions with no setup at all, and Threads started marking posts that sparked a real discussion.
Story of the day: LinkedIn explained how it finds and hides AI slop
On October 8, 2026 LinkedIn's vice president of engineering described how the platform finds AI slop and removes it from the feed. AI slop is what LinkedIn calls templated or repetitive posts with no real value, usually generated by a neural network. Social Media Today published the breakdown on October 9.
The starting point was this summer. LinkedIn added a "looks like AI slop" report option to posts. Users clearly had been waiting for it. In under 3 weeks after launch, more than 1 million people used the button. For a platform that rarely gets that kind of response to a reporting feature, this is a loud signal: people are tired of the same polished, empty posts.
Here is how the system works. User reports are used to train the models. On top of that, LinkedIn runs a separate set of AI agents, and each of them has its own rule. One agent looks for one pattern, another looks for a different one. Together they recognize such posts and remove them from the feed. In other words, a neural network now catches what other neural networks wrote, and people teach it what counts as junk.
Who needs to know this. Anyone who runs a company page or an expert profile on LinkedIn, and agencies that produce posts for clients in bulk. Until recently, a generated post was a cheap way to stay visible: publish every day, the algorithm notices activity. Now the math is different. A templated post does not just fail to get engagement, it loses impressions. And if readers start reporting your posts, that feeds the models too.
What to do: keep the neural network for drafts if it helps, but the example, the number and the opinion in the post have to be yours. Read last month's posts with a fresh eye. If they could be published by any company in your industry with the logo swapped, that is exactly the pattern LinkedIn is now filtering out. Fewer posts with your own experience will do more than daily filler.
More news on October 10, 2026
Threads is Meta's text-based social network. Until now the app had no way to highlight such posts. In August the feature was only spotted in the code during a test. Gems work as a daily selection of standout posts and replies. The algorithm picks them by 3 criteria: a fresh perspective, a lively discussion and originality. Only public posts qualify, and among replies only direct replies to the post count, a reply to a reply does not. Posts carried over from another app do not get the mark. Gems are awarded for the level of discussion, not for follower count, so a small profile has a chance too. The author gets a notification and can publish a special post about the mark. For now the feature works only for users in the US. For creators this is a direct signal: cross-posting from Instagram to Threads no longer helps, you need your own text, one that people want to argue with.
Grok Bot works inside X: you can give the agent a task, and it completes it in the background. Now it analyzes posts, finds trends, tracks brand mentions and suggests new opportunities. Before, starting August 29, 2026, you had to link the agent to your account through X connector, and only after that could the bot search posts, read the feed and check mentions. Now, according to X, no setup is needed. According to Crypto Briefing, the agent searches X directly and goes through more than 100 posts per request. Social Media Today reported the update. If your brand or channel is on X, this is free monitoring: you can ask for a weekly industry digest or keep an eye on product reviews without paying a separate service.
SynthID is an invisible watermark that Google DeepMind has been embedding in generated images, video and audio since 2023. Before, the detector was available only to the company's partners. Now anyone can use it. The service is free, works in the browser, and you can sign in with a Google, Apple or OpenAI account. The detector recognizes content from Google, OpenAI, Nvidia and Kakao models, and Apple will be added later. The watermark is already on more than 180 billion images and videos, plus audio recordings totaling 240,000 years. There is one important limitation: if the watermark was removed, the detector will show nothing. ppc.world wrote about the opening on October 9. If you make creatives and covers with neural networks, run them through the detector before publishing, so you know whether the AI label is visible to platforms and viewers.
The first test removes favicons from snippets. A favicon is the small site icon next to a search result, and right now it is how people recognize a familiar brand on the page. If it goes away, sites will look more alike, and recognition will rest only on the title and description. The second test shows the local pack without borders. The local pack is the map and list of businesses for queries like "coffee shop near me". The third test adds fading divider lines at the end of rows. For now these are experiments, not changes for everyone. If a noticeable share of your traffic comes from Bing, check how recognizable your page titles and descriptions are on their own, without the icon.
Orkut was especially popular in India and Brazil. Its creator, Orkut Büyükkökten, is calling on users to support a new social network built around friendship and communities, not around an algorithmic feed. The petition was signed by 2.73 lakh people, which is 273,000 in the usual count. The Indian outlet BestMediaInfo reported it. Right now this is a gauge of interest, not a product where you can create an account. But the number tells you something: more than 270,000 people are ready to sign up for a social network where communities come first. For creators this is an argument for owned channels that do not depend on a feed: group chats, newsletters, private groups.
What to do about it: creators, advertisers, site owners
- Creators and anyone running company profiles on LinkedIn and Threads: write posts people want to argue with, and do not copy them from other social networks. LinkedIn cuts templated AI posts, and Threads does not mark posts carried over from other apps.
- Brands on X: give Grok Bot the task of a weekly digest of mentions and reviews, it costs nothing.
- Anyone making creatives with neural networks: run images, video and audio through SynthID Detector before publishing to know whether the AI watermark is visible.
- Site owners with traffic from Bing: check that your titles and descriptions are recognizable without a favicon.
- Creators who depend on one platform: start building a channel the feed cannot take away, such as a newsletter or a private group.
Writes about: SMM, paid boosts, ads, traffic, suppliers, the market from the inside, daily news
