Which LinkedIn format, account and structure AI engines cite, engine by engine, with a 30-day plan.
LinkedIn AI citations go to a narrow slice of what companies publish there. In OtterlyAI's 30-day experiment, published November 9, 2026, the team posted 240 pieces on 60 topics. Articles from a personal profile earned 78 citations. Feed posts from that same profile earned zero.
Most B2B teams run LinkedIn the other way round: short posts from the company page, written for the feed. This guide covers which format, account and structure AI engines pull from LinkedIn, engine by engine, with data from seven 2026 studies. You'll also get a 30-day plan you can hand to three of your subject experts on Monday.
LinkedIn appears in a small share of all AI answers and a much bigger share of B2B ones. cloro found linkedin.com in 1.61% of 390,151 AI answers in the 30 days to August 7, 2026. That made it the 14th most-cited domain, far behind YouTube at 14.07% and Reddit at 11.30%. cloro's prompts skew toward retail, travel and consumer goods.
Business questions look different. Meltwater and LinkedIn analyzed 9.5 million citations across six AI platforms in May 2026. LinkedIn ranked first in Marketing & Advertising and in AI & Data Science, and third in Technology & SaaS, out of 14 business categories. Profound's March 2026 research ranked LinkedIn first for professional queries across six models.
The trend points up. OtterlyAI measured LinkedIn's share of social media citations at 7.8% in January 2026 and 11.7% in May.
Check your own category before you budget. Run 20 buyer prompts through each engine and count how many answers cite a linkedin.com URL.
Perplexity and Google AI Overviews cite LinkedIn most often, ChatGPT is climbing, and Gemini almost never cites it. cloro's October 2026 update covers 366,217 answers from the 30 days to October 12:
Two of those numbers moved a lot in two months. ChatGPT tripled from 0.57% in August. AI Mode fell from 6.20% after Google cut its sources per answer from about 14 to about 5 at the end of August, according to cloro.
Gemini is a dead end for now. OtterlyAI counted 38 Gemini citations among 1.31 million LinkedIn citations between January and June 2026.
Pick the one or two engines your buyers use before you plan content, because the format that wins changes by engine. Engines share few sources with each other, and LinkedIn follows the same pattern.
Articles earn more citations in every large 2026 dataset, though ChatGPT and Google AI Overviews now lean toward feed posts. Goodie's October 30, 2026 study counted 219,351 citations across nine engines from January through September. Articles took 171,762 of them and feed posts took 47,589, a 3.6x gap. OtterlyAI's January to June data shows the same thing per URL: 8.5 citations for an article against 5.9 for a post.
The mix is shifting, though. In cloro's October data, 51.8% of LinkedIn-citing answers cited a feed post and 35.9% cited an article. In August those figures were 33.3% and 59.1%.
The split depends on the engine. On OtterlyAI's November prompt set, feed posts made up 66.4% of ChatGPT's LinkedIn citations and 58.9% of AI Overviews'. Copilot cited articles 100% of the time and Claude 91.0%.
So lead with the article, then publish a 150 to 250 word post that answers the same question. Monitor your progress with Nobori: tag each tracked prompt with the format the engine cited.
Publish from a named person. Meltwater found that 75% of LinkedIn citations came from individual members and 25% from Company Pages. OtterlyAI's experiment put a number on the gap for articles: 1.30 citations per piece from a personal profile against 0.45 from the company page, or 2.9x. Engines cited 19 of the 60 personal articles and 9 of the 60 company ones.
Company pages still pull weight on long-form. Goodie split article citations 51% personal and 49% company, close to even. Feed posts went 73% to personal profiles, and company pages led both formats in just 3 of 32 industries.
Short company-page posts are the weak spot. OtterlyAI's 60 company posts earned 2 citations between them.
Here's how I'd split the work. Move feed posting to your experts. Keep the company page for articles that belong to the brand, like research reports and product explainers. If you sell into Google-heavy categories, weight personal posts higher: Goodie found 86% of AI Overviews' feed post citations went to personal profiles.
No. OtterlyAI ran correlations across 161,440 cited LinkedIn URLs and found none worth acting on: likes at r = -0.06, comments at -0.04, hashtags at -0.02. The median cited piece had 8 likes.
The experiment makes the point harder to ignore. Personal-profile articles got a median of 2 views and 1 reaction, and they won 72.9% of the campaign's citations. Company-page articles got a median of 120 impressions and 4 reactions.
Semrush saw the same thing in March 2026 across 89,000 cited LinkedIn URLs. The median cited post had 15 to 25 reactions and at most one comment. Creators with fewer than 500 followers got cited as often as bigger accounts.
One counterpoint: LinkedIn's Brooke Weller said at brightonSEO San Diego in September 2026 that cited content tends to earn 10 or more thoughtful comments, per OtterlyAI's write-up. None of the public datasets back that up yet.
Stop reporting likes as your AI metric. Report citations per URL.
Use the buyer's exact question as the title and answer it in the first two sentences. In OtterlyAI's experiment, 87.9% to 95.3% of campaign citations came from prompts that matched the article title.
Meltwater looked at the 24 most-cited LinkedIn articles in its dataset. All 24 used bullets or numbered lists. 92% used H2 or H3 headings, 75% named specific companies or tools, 67% included hard numbers, and 50% had a comparison framework.
That gives you a brief for every article:
On feed posts, skip generic hashtags. OtterlyAI found the hashtag ends up in the post URL, and 75.8% of its 120 test posts shared the slug "geo-aisearch". Without a hashtag, your first line sets the slug.
The same rules apply on your own blog. See the chunk-first framework for the long version.
Aim for 800 to 1,200 words. LinkedIn's own guide recommends that range, as Social Media Today reported in March 2026. The cited-content data lands in the same place: OtterlyAI's median cited article ran 1,021 words between January and June 2026, and 1,142 on its November prompt set.
Semrush puts the band wider, at 500 to 2,000 words for articles and 50 to 299 words for feed posts. OtterlyAI's median cited post was 185 words.
Length buys you nothing past the point where the question is answered. OtterlyAI's correlation between word count and citations was r = 0.03. Write until a buyer could act on the answer, then stop.
Engines cite content that helps a buyer decide: lists, comparisons and how-to-choose guides. In Meltwater's May 2026 data, "Best X" lists made up 54% of cited LinkedIn content, side-by-side comparisons 50% and how-to-choose guides 33%. Educational explainers came in at 17% and thought leadership with supporting data at 8%. A piece can sit in more than one group.
Original work wins. Semrush found about 95% of cited posts were original and 5% were reshares. It also found that 54% to 64% of cited posts gave knowledge or advice.
Fresh work wins too. Meltwater found 48% of cited posts were under three months old and 12% were over a year old.
Build your topic list from prompts. Meltwater suggests 25 to 50 high-intent prompts covering best, compare, how to choose, pricing, risks and implementation. Go narrow: "best SOC 2 automation tools for a 50-person SaaS" beats "compliance trends". Our guide to comparison content for AI search covers how to write these without ranking yourself first on every list.
Expect a first citation within about a week on Copilot or Perplexity, and a wait of months on Google and ChatGPT. In OtterlyAI's experiment, cited pieces took a median of 7 days from publish to first citation. The slowest took 28.
The engines moved at different speeds. Copilot cited a campaign article on October 2, four days after the first piece went live. Perplexity followed on October 13. Of 107 campaign citations in the 30-day window, Copilot gave 81, Perplexity 25 and AI Mode 1. ChatGPT, AI Overviews, Claude and Gemini gave none. ChatGPT cited two of the articles between November 6 and 8, after the window closed.
Google is slow because incumbents hold the slots. Six LinkedIn URLs from outside the test took 77.2% of the LinkedIn citations in AI Overviews and AI Mode on OtterlyAI's prompts.
Don't use Google rankings as your scorecard here. None of the 240 pieces ranked for a tracked prompt, and 82.4% of AI Overview citations to LinkedIn went to URLs that didn't rank.
Monitor your progress with Nobori: watch Perplexity for the early signal and give Google a quarter. Copilot reads from Bing, so your Bing setup counts as well.
Keep your own site first. In cloro's August 2026 data, a page on the brand's own domain appeared in 19.86% of AI answers. A LinkedIn article appeared in 1.15%. That's about 17x in favor of owned pages. Third-party listicles appeared in 43.15% of answers.
LinkedIn's share also swings. On a fixed set of 116 prompts, cloro saw LinkedIn fall from 13.80% of answers in February 2026 to 4.83% in July, a 65% drop. Reddit fell 40% and Wikipedia 92% over the same months, as engines cut average sources per answer from 13.95 to 11.19.
The cost side is small. OtterlyAI spent 35 hours across one month to publish 240 pieces. Three experts writing two articles a month is a fraction of that.
I'd treat LinkedIn as a second surface for work you've already done. Take your best explainers and comparisons, rewrite each one under an expert's name, and publish it as an article. Then put the larger share of off-site effort into third-party mentions and roundups.
Three experts, six articles and one tracked prompt list fit into 30 days.
Posting rhythm counts. Semrush found about 75% of cited post authors had published five or more posts in the previous four weeks.
Monitor your progress with Nobori: run the same prompt list every day and compare week 4 against your week 1 baseline.
Track citations per LinkedIn URL, per engine, on a fixed prompt list. Weller's suggestion, relayed by OtterlyAI, is 7 to 10 prompts per topic. For each cited URL, log five things: the prompt, the engine, the URL, the format (article or post) and the account (person or company).
Expect a lopsided result. In OtterlyAI's January to June 2026 data, the top 1% of cited LinkedIn URLs took 30.2% of all citations, and the median cited URL got two. 87.4% of cited URLs appeared on one engine. The 2.6% that appeared on three or more engines captured 24.0% of citations.
So judge the program by its best two or three pieces. One article that four engines cite is worth more than twenty that get one citation each.
Log model releases next to your numbers. OtterlyAI saw ChatGPT's LinkedIn share move from about 0.1% to about 0.4% of citations on August 7, 2026, the day of a model update. We covered that pattern in ChatGPT citation drops.
Every study here counts cited URLs and nothing else, and the vendors disagree on basics. Semrush says articles make up 50% to 66% of cited LinkedIn content. Meltwater says 12%, with text posts at 72%. Profound put long-form articles at 8.9% of ChatGPT's LinkedIn citations in February 2026, counting profiles in the total. The prompt sets, engines, dates and definitions all differ.
The OtterlyAI experiment is one test. It used one personal account with 480 followers, one subject (GEO) and 179 US prompts. OtterlyAI says the results describe patterns on its own prompts. Goodie estimates a 4.6% error rate in its personal-versus-company labels.
Every source also sells an AI visibility or content product. So do we.
Where the studies agree, I'd trust them: named people beat company pages on feed posts, engagement doesn't predict citations, and question-titled articles with lists get quoted. For the rest, run your own 30-day test before you move headcount.
Nobori tracks your prompts across ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude, and refreshes the data daily. For each prompt you see which sources the engine cited and which brands it named. That tells you where LinkedIn shows up in your category and where it doesn't, before you ask anyone to write.
You also see who gets cited where you're missing. If a competitor's head of product owns the LinkedIn slot for "how to choose" prompts in your category, Nobori shows you the prompt and the cited page. It then turns the gap into a task your team can assign.
Once your experts start publishing, the same prompt list gives you the before and after. You can compare week 1 with week 4 by engine and keep the program honest about what earned a citation.
See if AI engines are citing you → nobori.ai