What I learned working with a technology advisory firm whose clients were turning to AI for answers, and what happened when we stopped posting more and started answering better.

If your LinkedIn strategy for small business has felt like a guessing game in 2025, what I am about to share will reframe it completely. I have been sitting with this one for a few weeks because I wanted to make sure the results were real before I wrote about it.

A client of mine runs a technology advisory practice. LinkedIn is their only social channel. Their audience is sophisticated, time-poor, and not scrolling LinkedIn for inspiration. They are on LinkedIn when they have a specific professional problem to solve and they want an answer from someone who actually knows what they are talking about.

We tried the standard advice first. Post more consistently. Improve the hooks. Mix up the formats. The numbers told us very clearly that it was not working. The average post was landing at 198 impressions, a 2% engagement rate, and 3 reactions. We were visible to almost nobody.

Then I changed the approach entirely. Not the posting frequency. Not the captions. The LinkedIn strategy for small business. And the results were different enough that I want to walk you through exactly what we did and why it worked, because the reason it worked has everything to do with how AI tools like Claude are now finding and citing content.

The LinkedIn strategy most small businesses have not caught up to yet

When your future clients have a question, many of them are no longer typing it into Google first. They are asking Claude, ChatGPT, or Perplexity. And those tools are pulling answers from somewhere.

LinkedIn is now one of the most cited professional sources across all major AI platforms. A Semrush analysis of 325,000 unique prompts across Claude, ChatGPT, Google AI Mode, and Perplexity found LinkedIn URLs in 14.3% of ChatGPT responses, putting LinkedIn ahead of Wikipedia, YouTube, and every major news outlet for professional queries. The citation rate more than doubled between November 2025 and February 2026.

That means a well-written LinkedIn article answering a specific professional question is no longer just reaching your followers. It is potentially being served as the answer the next time someone asks an AI tool about that topic. That is a fundamentally different kind of visibility than a post that performs well in the feed for 48 hours and then disappears.

LinkedIn Pulse Articles account for approximately 73% of all LinkedIn citations in AI responses, while regular feed posts account for around 10.5%. And 95% of cited content is original. Reshares are almost never cited.

Your LinkedIn article is no longer just content for your followers. It is a potential answer to a question your future client is asking Claude right now.

This is why getting cited by AI tools has become one of the most important visibility strategies for service businesses in 2026. And LinkedIn, with its professional data layer and citation rate, is currently the most direct path to it.

Why LinkedIn changed the rules at the same time

In May 2026, LinkedIn officially confirmed an algorithmic overhaul that had been rolling out for months. The platform also launched two new post metrics, Saves and Sends, at the same time. Those two things together tell you exactly what a winning LinkedIn strategy for small business looks like right now.

The old LinkedIn distributed your content to your network. The new LinkedIn distributes your content based on interests. The platform now builds an interest graph for every user, and routes content to people based on what they have been engaging with professionally, not who they follow. A post from an account with 400 connections can now reach thousands of people in a specific industry if the content matches their interest profile.

The Saves and Sends metrics matter because they are the exact signals the algorithm uses to evaluate whether your content is genuinely useful. A Save means someone thought your post was worth coming back to. A Send means someone forwarded it privately to a colleague. Both signals tell the interest graph that your content has real value beyond the initial impression. LinkedIn is literally showing you the ranking signals in your analytics dashboard now.

The formats earning the highest engagement under the new system: document posts (PDF carousels) at 6.60% average engagement rate, and native video at 5.60%. Both significantly outperform standard text posts. Both require something most generic content does not have: a specific answer, from a specific expert, about a specific thing.

What the algorithm suppresses: formulaic content that could have come from any business in any industry. Posts that use AI to generate marketing-sounding output with no real expertise behind it. Engagement bait. Generic advice. The system has become very good at detecting the difference between content that sounds expert and content that actually is. This is where most LinkedIn strategies for small businesses break down.

What we did differently and what the numbers showed

Back to my client. The technology advisory firm with 198 impressions per post and 3 reactions.

The first thing I changed was the research process. Instead of asking what should we post about this week, I used Claude to map what questions their audience was actually asking. Not general questions about the industry. Specific, urgent questions tied to real problems their clients were navigating right now.

What came out of that research surprised even my client. There was a Canadian regulatory deadline coming in September 2026 that many of their clients were required to comply with. The deadline was real, the stakes were significant, and a large number of their clients were uncertain about where they stood and not talking about it publicly. People were asking AI tools about it. They were searching for plain-language explanations of what compliance actually required. And almost nobody in the advisory space was creating content that answered those questions directly.

We used Claude to dig into the specific questions being asked, the language being used, and the gaps in the existing content landscape around that regulatory topic. Then we built content around those answers. Short, specific, answer-first videos. Not here is our perspective on the regulation. Here is the question your team is avoiding. Here is what readiness actually looks like. Here is where most organisations have the gap.

The results over three weeks:

• Average post before: 198 impressions / 3 reactions / 2% engagement
• Video content after: 1,273 impressions / 49 reactions / 4% engagement / 826 reach / 527 unique viewers / 128 minutes of watch time / 7 comments
• And inbound reach out from the right people.

The impressions went up more than six times. The reactions went up more than sixteen times. And the reach extended well beyond the existing follower base, to people in the interest graph who matched the topic profile, including people who had never encountered this firm before.

The watch time matters most to me. 128 minutes of watch time across a three-week period means people were not just scrolling past. They were stopping, watching, and in several cases, reaching out directly. That is not a content win. That is a business development outcome from a LinkedIn strategy for small business that started with research, not a content calendar.

The process: how we used Claude to find the right questions

This is the part I want to be specific about because it is replicable for any service business with a defined niche and a professional audience on LinkedIn.

Step one is not writing anything. Step one is research. The question I asked Claude was not write me some LinkedIn posts. It was: what are the specific questions people in this industry are asking about this topic right now? What language are they using? What are they uncertain about? What is the gap between what they know they need to do and what they are actually asking about?

A detailed, structured prompt that includes your client’s industry, their audience’s role and seniority, the specific topic or challenge, and the outcome they are trying to reach will produce research that a generic prompt simply cannot. This post on prompt structure covers exactly how to build that kind of prompt. The difference between a surface-level response and a genuinely useful research output is almost always in the specificity of what you give the AI to work with.

Step two is gap analysis. Once you have the questions, the next prompt asks Claude to identify what content already exists that answers those questions, and where the gaps are. For the regulatory topic we were working on, the gap was significant. There was plenty of formal documentation. There was almost no plain-language, practitioner-focused content that answered the questions their audience was actually asking.

Step three is format selection. Not every question is a video. Not every answer is an article. Short, urgent, answer-first questions work well as videos because the viewer gets the answer fast and the watch time signals tell the algorithm it is relevant. Deeper explanations of process or framework work well as articles because they earn saves and get cited by AI tools. The format follows the question, not the other way around.

Step four is the brand foundation check. This is the step most people skip and it is the reason AI-generated content gets suppressed by LinkedIn’s algorithm. Every piece of content we produced was reviewed against the client’s actual voice, their specific expertise, and the language their clients use. When AI generates content from a documented brand foundation, the output reads as expert. When it generates from a generic prompt with no brand context, it reads as marketing. The algorithm knows the difference. So do your clients.

Why we are now adding LinkedIn articles alongside the blog

Here is the strategic shift I am making with this client going forward, and it is directly connected to the AI citation data. That is what a good LinkedIn strategy for small business looks like in 2026: one piece of content, two publications, four jobs.

The blog on their website is the foundation. It builds domain authority, it is indexed by Google, and it is where the canonical content lives. But LinkedIn Pulse Articles, the long-form format published directly on LinkedIn, are where the AI citation opportunity is. Seventy-three percent of LinkedIn citations in AI responses come from articles, not feed posts.

The workflow is straightforward: publish on the website first, wait three to five days for Google to index it as the original source, then publish the full article on LinkedIn with a link back to the original. Slightly tweak the LinkedIn headline to target a marginally different search phrase. Add a line at the top of the LinkedIn article noting where it was originally published.

This approach does not create a duplicate content penalty. Google will generally index the website version as the original and the LinkedIn version as a syndication. The LinkedIn article, because it lives on a domain with high professional authority and citation rate, becomes the version that AI tools are most likely to surface when someone asks a relevant question.

One piece of content. Two publications. Four jobs: feed visibility, domain authority, Google indexing, and AI citation.

What this means if LinkedIn is one of your channels

If your audience is professional, your niche is specific, and you have been posting on LinkedIn without seeing meaningful results, the problem is almost certainly not the frequency or the format. It is the questions you are answering. That is the foundation of any LinkedIn strategy for small business that actually compounds.

Generic content gets distributed generically. Specific content, content that answers a real question your specific audience is actually asking right now, gets routed by the interest graph to the people who are already looking for that answer. And if it is written as a LinkedIn article with enough depth and originality, it also gets cited when those people ask an AI tool the same question.

The starting point is not a content calendar. It is a research session. Use Claude to map the questions your audience is asking. Look for the gaps. Find the urgent, specific, underserved question that your expertise actually answers. Build your content around that answer. Then publish it in a format the algorithm rewards and in a place AI tools are already looking.

That is a different kind of content strategy than posting consistently and hoping for reach. It is slower to set up. It compounds faster once it is running.

The bottom line

LinkedIn changed how it distributes content in 2026. At the same time, it became one of the most cited professional sources across Claude, ChatGPT, and Perplexity. Those two facts together mean that a single well-researched LinkedIn article, answering a specific question your audience is actually asking, now has more potential reach than a month of consistent feed posts.

The technology advisory firm I worked with went from 198 impressions and 3 reactions per post to 1,273 impressions, 49 reactions, and real inbound reach out over three weeks. Not by posting more. By answering better, in the right format, from a documented brand foundation, after doing the research to know what questions actually needed answering.

Claude did the research. The client provided the expertise. The combination produced content that the algorithm rewarded and that AI tools are now in a position to cite. A LinkedIn strategy for small business in 2026 has two jobs: feed visibility and AI citation. This is how you do both at once.

Key takeaways

• LinkedIn is cited in 14.3% of ChatGPT responses and around 11% across all major AI platforms. LinkedIn Pulse Articles account for 73% of those citations. Feed posts account for 10.5%.
• LinkedIn officially confirmed in May 2026 that its feed now runs on interest graphs, not connections. Specific, expert content in a defined niche travels further than generic content regardless of follower count.
• LinkedIn launched Saves and Sends as trackable post metrics in May 2026. These are the exact signals the algorithm uses to evaluate genuine content value.
• Short, answer-first videos and long-form articles are the two highest-leverage formats. Videos earn reach and watch time signals. Articles earn saves and AI citations.
• The research step is the most important step. Using Claude to map what questions your audience is actually asking, in their language, about their specific urgent problems, is what produces content that gets distributed and cited.
• Publishing to your website first and LinkedIn three to five days later, with a link back, captures both Google indexing and AI citation without a duplicate content penalty.
• The fix is always the foundation. AI-generated content that draws from a documented brand voice and real expertise passes LinkedIn’s quality filter. Generic AI output does not.

Frequently asked questions

Will republishing my blog post as a LinkedIn article hurt my Google ranking?

No, as long as you publish on your website first and wait three to five days before posting to LinkedIn. Google will index your website as the original source. Include a link back to your original post at the top of the LinkedIn article. Your website retains the SEO authority and the LinkedIn article does the citation work. Two publications, one piece of content, no penalty.

How do I find the right questions to answer for my LinkedIn audience?

Start with Claude. Give it your industry, your audience’s role and level of seniority, the specific topic or challenge you work in, and ask it to surface the questions people in that field are actively asking. Ask for the language they use, the gaps in existing content, and the questions that are urgent but underserved. The more specific your prompt, the more useful the research. A detailed brand foundation document fed into the prompt significantly improves the output.

Does it matter what type of LinkedIn content gets cited by AI?

Yes significantly. LinkedIn Pulse Articles (long-form content published directly on LinkedIn) account for 73% of all LinkedIn citations in AI-generated responses. Regular feed posts account for around 10.5%. Original content is cited 95% of the time. Reshares almost never appear. The content that gets cited answers a real question clearly, is written by an identifiable expert, and is specific enough to be genuinely useful to someone asking about that topic.

How long before the LinkedIn interest graph starts routing content to the right people?

Most analysis points to 60 to 90 days of consistent, topically focused posting before the interest graph reliably categorises your expertise and routes content to the right audience. Aligning your profile headline and About section with your content focus at the same time speeds this up. The regulatory content we produced for the technology advisory client started seeing meaningful reach within three weeks because the topic was highly specific and the audience interest signal was already strong.

Do I need a large LinkedIn following for this LinkedIn strategy for small business to work?

No. The interest graph distributes content based on topic relevance, not follower count. The technology advisory firm we worked with did not have a large following. What they had was specific expertise, a specific audience with a specific urgent problem, and content that answered that problem directly. The algorithm routed it to the right people because the semantic match was strong. Follower count is no longer the ceiling on LinkedIn reach. Specificity is the mechanism.

Ready to build a LinkedIn strategy that actually compounds?

The AI Blueprint Prompt Library includes a complete set of research prompts built for service businesses, structured to help you map audience questions, identify content gaps, and build a LinkedIn strategy for small business from the answers up. There is also a Custom GPT that builds your brand foundation in about 20 minutes, so every piece of content you produce, with or without AI, draws from the same documented voice.

Two paths to the same place. The Prompt Library at CA$25.99 one-time if you want to build it yourself. The AI Clarity Kit at CA$499 if you want to build it with guidance. Full access from day one. No subscriptions.