LinkedIn algorithm 2026: what changed, and the AI data setting to check

The LinkedIn algorithm in 2026 runs on a feed that LinkedIn rebuilt in March around large language models, the technology behind AI assistants. It works out what each post is about and shows it to members interested in that subject, including people outside your network. LinkedIn also said it would show less engagement bait and act against automated comments.

The second change concerns your data rather than your reach. Since 3 November 2025, LinkedIn has used the profiles and public posts of members in the UK and Europe to train its generative AI, unless they switch it off. It takes a minute to check, and I show you where below.

I do not sell social media management. When I helped a client with social media, posting and joining conversations brought them 60 relevant new followers in a month, but I prefer work whose results show up as enquiries. What follows is a dated list of what LinkedIn changed, checked on 17 October 2026, and what each change means if you sell to other businesses. To compare LinkedIn with the other networks, see my guide to the best social media for small business.

LinkedIn algorithm 2026: what changed in March

On 12 March 2026, LinkedIn’s engineering team described a new feed system on the LinkedIn Engineering blog, and the trade press, including Search Engine Land, covered the main points. A feed works in two steps, and LinkedIn rebuilt both:

1. Retrieval picks the posts that might interest a member. It now relies on embeddings produced by a large language model (numerical summaries of what each post means), so it can connect related subjects. In principle, a post about air conditioning for restaurant kitchens can reach people interested in restaurant fit-outs, even if they have never heard of you.
2. Ranking puts those posts in order. A sequential model, from the same transformer family as AI assistants, learns from each member’s history of likes, comments and time spent reading to predict what they will engage with next.

In the same update, LinkedIn said it would show less content built only to drive comments or clicks: posts asking you to comment “YES” (often to receive a guide), videos paired with unrelated text, and recycled thought leadership. It also said it acts against comment automation tools, browser extensions and engagement pods, the groups whose members agree to like and comment on each other’s posts.

For a small business, the plain meaning is this: what your post is about now matters as much as who you know. A clear, specific post can travel beyond your connections to people who care about its subject. A vague one gives the system nothing to match.

Why your reach may have dropped

Several third-party studies have reported lower average reach on LinkedIn in 2025 and 2026. Their figures disagree widely, from under 10 per cent to more than half, and LinkedIn has not confirmed any of them, so I would not plan around a number.

The official changes explain two things. If your posts relied on bait, pods or automated comments, they have lost that help. If they were vague, they now compete with posts the system understands better.

LinkedIn changes since 2024, in date order

The AI data setting to check: Data for Generative AI Improvement

Since 3 November 2025, LinkedIn has used data from members in the UK, the EU, the EEA and Switzerland, as well as Canada and Hong Kong, to train the AI models behind its content-creation features. That covers profile details and public content such as your posts. LinkedIn says it does not use private messages, InMail included, and it leaves out members under 18.

LinkedIn relies on legitimate interest rather than your consent, so the setting stays on unless you switch it off. The UK GDPR and the EU GDPR give you a right to object, which this switch lets you exercise:

1. On a computer, click your photo (Me) at the top of LinkedIn, then Settings & Privacy. In the app, tap your photo, then open the settings.
2. Open Data privacy.
3. Under How LinkedIn uses your data, select Data for Generative AI Improvement.
4. Switch off Use my data for training content creation AI models if you do not want your data used.

If a label has moved, LinkedIn’s help page on this setting has the current wording. Switching off stops future use, but it does not undo training that has already happened.

Each member has their own switch, so if colleagues post for your business from their profiles, ask them to check theirs. Whether to switch it off is your call: I use AI tools in my own work, and I simply want you to choose rather than inherit a default. The same question applies the other way round, to what client data you can put into an AI assistant.

Personal profile or company page?

Post from both, but put most of your effort into your own profile. LinkedIn is built around members: people connect with and follow people, while a company page mainly reaches those who follow it.

Buyers also want to know who they will deal with. In the cold email campaigns I run for an air conditioning installer near Paris, many architects called instead of replying: a call is quicker, and they like real human contact, as I describe in this cold email case study. On LinkedIn too, a name and a face earn more trust than a logo.

For a business of one to ten people, post under your own name and photo, and treat the page as your shop window. When a prospect looks you up, the page should confirm what your profile says: same services, same website, same tone.

What a B2B small business should post on LinkedIn in 2026

The March rebuild rewards posts that the system can understand and that people actually read. That suits small businesses: you know your trade better than any influencer does.

1. One subject, stated plainly. Name your trade, your clients and their problem in the first two lines. “Why office air conditioning fails in the first heatwave, and how to check yours” gives the system something to match; “Monday motivation” does not.
2. Real work, shown. Photos of a job, a before and after, a quote you had to explain, a mistake you often see clients make.
3. One question per post. The questions you hear on the phone every week are your best topics. People spend time reading the answer to their own problem, and time spent reading is one of the signals the ranking model learns from.
4. No bait. No “comment YES and I’ll send you the guide”, no unrelated video glued to a text post, no recycled thought leadership. LinkedIn named all three in March.
5. Comments written by hand. Leave real comments on posts in your field, without pods or automation tools. Since August, LinkedIn sorts comments by their relevance to each reader, so a useful comment can appear near the top for the readers it concerns.
6. A reply to every comment on your posts, with an answer that adds something. LinkedIn’s August change was designed to encourage exactly these conversations.
7. A next step. End with your phone number or your website for people who want more, the way a good cold email ends.

Is LinkedIn worth your time in 2026?

If you sell to other businesses, yes, as long as you treat it as a long game. LinkedIn is where your buyers are in work mode, and a clear profile with useful posts makes you easier to trust when they look you up. If you are thinking of a LinkedIn newsletter, my comparison of LinkedIn newsletters, email newsletters and Substack shows who keeps the subscribers on each.

But do not add LinkedIn on top of five other channels. It is easy to fall for every new tool and platform; what works is to focus on the channels you already run and master one before adding the next, as I explain in shiny object syndrome. If your customers are households rather than companies, LinkedIn is probably not that channel.

Paid reach is a separate decision. LinkedIn Ads burn through cash fast, but done right, the clients they bring are bigger tickets; my LinkedIn Ads tutorial shows how to keep the budget under control.

And do not chase followers for credibility. “I have 10k followers” impresses less than “this magazine published an article about my business”, and the second needs no followers at all: see how to get press coverage.

Your LinkedIn checklist for the next 30 days

1. Check Data for Generative AI Improvement in Settings & Privacy, then Data privacy, and ask anyone who posts for your business to check theirs.
2. Rewrite your headline: what you do, for whom and where. “Air conditioning for restaurants and offices in Leeds” says more than “Founder and CEO”. Then make your company page say the same.
3. List ten questions clients ask you on the phone, and turn four of them into posts, one a week.
4. Clean up your habits: remove the bait from any post templates you reuse, leave any engagement pod or automation tool, and reply to every comment within a day.
5. Add your website to your profile and your page with a tracking link, so LinkedIn visits show up in your analytics:

https://www.yourbusiness.co.uk/?utm_source=linkedin&utm_medium=social&utm_campaign=profile

6. Review after 30 days: profile views, website visits from LinkedIn and enquiries. If nothing has moved and your buyers are not on LinkedIn, give those hours to a channel where they are.