AI visibility is still a new discipline, and that was clear from the questions we received during our recent webinar with LinkedIn.
We heard questions about almost every part of the process; What makes an AI model see someone as an expert? Should brands prioritize personal profiles or Company Pages? Do boosted posts help? Are articles more valuable than feed posts? How should you measure success? And what happens when the sources AI models prefer change?
Our original research with LinkedIn analyzed 9.5 million citations across six major AI models and 16 B2B categories. The findings gave us a useful picture of what AI systems were citing, but they also raised a bigger set of practical questions about how brands should respond.
Here, we answer the questions that came up most often.
Contents
What does it actually mean when we say AI “cites LinkedIn”?
What makes AI treat someone on LinkedIn as an expert?
Should brands prioritize personal profiles or Company Pages?
Are LinkedIn posts or articles better for AI visibility?
Does word count affect your chances of being cited?
Do boosting, engagement, tags, and links improve AI citations?
What about images and video?
How should brands work with employees without controlling what they say?
Can AI amplify bad statistics or misinformation from an “expert”?
How should you benchmark your current AI visibility?
What does success look like for executive LinkedIn content?
How should we think about reposting the same content from company and personal profiles?
What about comments and participation in trending conversations?
How much should brands invest when AI source behavior changes so quickly?
Does this research apply to nonprofits or the public sector?
What should brands do next?
FAQ
What does it actually mean when we say AI “cites LinkedIn”?
When we talk about an AI citation in our research, we mean a domain or URL referenced by an AI platform in a generated response.
Many AI experiences now actively retrieve information from the web and show the sources used to support an answer. Google AI Mode, for example, searches across web sources and provides links alongside generated responses. ChatGPT can search the web and provide cited answers, while Claude’s web search similarly provides direct citations and source links.
Our study tracked citations exposed in those kinds of responses across Copilot, Google AI Mode, Google AI Overviews, Claude, ChatGPT, and Gemini.
That also explains why citation behavior varies between platforms. Each product retrieves, selects, ranks, and presents sources differently.
For marketers, the practical question is therefore broader than “Does ChatGPT know about us?” Ask instead: when someone asks an AI platform an important question about our category, do we appear in the answer, how are we described, and which sources support that description?
What makes AI treat someone on LinkedIn as an expert?
There is no public checklist showing exactly how every AI platform assigns authority to an individual.
Our data does, however, show some clear patterns. Individual members accounted for 75% of the LinkedIn citations we studied, compared with 25% from Company Pages. Citation volume also did not simply follow audience size. More than half of citations came from members with fewer than 10,000 followers.
Profile context appears relevant too. LinkedIn provides signals such as a member’s job title, employer, industry, and publishing history. In our dataset, senior leaders including CEOs, engineering VPs, CPOs, and CTOs featured prominently among cited authors. That does not mean a CEO title automatically creates authority, or that an analyst, professor, engineer, or practitioner is less valuable.
A better rule is to match the person to the topic.
A product engineer explaining a technical implementation may be more credible than a CEO discussing the same subject at a high level. A financial analyst may have stronger authority on valuation than a general marketing executive. Expertise should come from proximity to the subject, supported by evidence and practical experience.
So do not rewrite organizational titles simply to influence AI, instead identify the people who genuinely know the topic you want your brand to be associated with, and help them turn that knowledge into useful public content.
Should brands prioritize personal profiles or Company Pages?
The research found that individual profiles generated three quarters of LinkedIn citations, but Company Pages still accounted for a meaningful 25%. The two sources also serve different purposes.
Company content is valuable for authoritative information about products, announcements, company facts, research, and official positions. Individual experts can add experience, interpretation, practical advice, and a point of view.
That creates a stronger strategy than choosing one over the other.
For example, a company might publish original research on its website and Company Page. A subject-matter expert can then explain what the findings mean in practice, while an executive discusses the strategic implications. A customer or partner might contribute another independent perspective.
The same principle applies when employees are reluctant to create completely original posts. Adding substantive personal commentary to company content can still be useful, but our research found that 72% of cited LinkedIn content was original, compared with 28% reshared content. That suggests brands should encourage experts to add genuine knowledge rather than simply repost corporate material.
Are LinkedIn posts or articles better for AI visibility?
This was one of the most common questions, and the answer needs some nuance.
Text posts accounted for 72% of the cited formats in our analysis, compared with 12% for articles, 11% for video, 5% for documents, and 1% for images. But when we compared performance at the individual-content level, long-form articles earned 6.5 times more AI citations than standard posts.
There are far more feed posts published than long-form articles, so posts can contribute more citations overall. An individual well-structured article, however, can provide much more material for an AI system to extract and cite.
Use posts for regular expertise, timely observations, data points, and concise answers. Use articles for subjects that need depth: comparisons, buying guides, frameworks, detailed explainers, or “how to choose” content.
LinkedIn newsletters can also fit this strategy because newsletter editions are LinkedIn articles published as part of a recurring series. LinkedIn says newsletters are discoverable, readable, and shareable both on and off the platform.
Our study did not isolate the newsletter feature itself as an AI ranking factor, so we would not claim that starting a newsletter automatically improves GEO. Its value comes from creating a regular home for substantive, structured content.
Does word count affect your chances of being cited?
Among the highest-performing LinkedIn articles we analyzed, 1,500 to 2,500 words was a common range, with a median of 1,725 words. Treat that as an observed pattern, not a target you need to hit.
An article does not become more citable because it reaches 1,500 words. The stronger signals in the research were usefulness and structure. Every one of the top-cited articles analyzed used lists. Ninety-two percent used clear section headings, 75% named specific companies or tools, and 67% included hard numbers or data.
Write for completeness. If a question can be answered properly in 800 words, adding another 700 words of filler will not improve it. If the subject requires 2,000 words to compare options, explain criteria, provide evidence, and answer follow-up questions, the additional depth can make the content more useful.
Do boosting, engagement, tags, and links improve AI citations?
We did not test paid boosting as part of the study, so there is currently no basis in our research for saying that ad spend directly increases AI citation likelihood.
LinkedIn describes boosting as a way to turn an organic post into an ad and increase its reach, impressions, engagement, followers, or website traffic. That can clearly improve human distribution, although the connection to AI visibility is less direct.
One of the more interesting findings from our research was that highly cited content did not necessarily have huge engagement. Detailed expertise could earn AI citations even when a post had relatively modest social performance.
The same caution applies to tagging, as we did not find evidence that mentioning more people or brands makes content more citable.
Links are more useful when they strengthen the information itself, because attribution gives readers, and potentially retrieval systems, clearer evidence to work with. If you quote a statistic, link to the original research. If you refer to a framework, identify its source. Avoid turning those practices into mechanical GEO tactics. Accurate sourcing is more valuable than adding tags or links simply because you hope an algorithm will notice them.
What about images and video?
AI-cited LinkedIn content is not limited to articles and text posts. Video represented 11% of the formats cited in our research. But brands should avoid putting the entire argument inside an image, chart, or video without useful accompanying text.
Make the core idea explicit in the post itself. State the finding, identify the entities involved, include important numbers, and explain the conclusion. For video, provide enough written context for someone to understand what the content covers before watching it.
This improves accessibility for people as well as making the content easier for retrieval systems to interpret.
How should brands work with employees without controlling what they say?
Several webinar attendees asked some version of this question, and the answer is enablement, not scripting.
Start by identifying the topics where the company has genuine expertise and the people who know those subjects best. Give those experts access to useful raw material: research, approved statistics, customer questions, product information, industry data, and source links, then let them add their own interpretation.
A useful employee program might provide a monthly briefing containing three customer questions, two original data points, a new product insight, and relevant external research. The employee decides what they think about those subjects and how they want to communicate it.
The same principle applies to disclosures. We did not test whether employee disclosures in a bio or post affect citation rates. There is no reason to remove appropriate disclosures in pursuit of GEO. Transparency and accurate attribution support credibility, and those should take precedence over speculative algorithmic advantages.
Can AI amplify bad statistics or misinformation from an “expert”?
Yes, this risk exists. Authority signals do not guarantee factual accuracy. An executive can quote a bad statistic. A respected publication can make an error. An AI system can misunderstand a source.
AI providers themselves make this limitation clear. OpenAI advises users to verify important information and cited sources, while Anthropic notes that Claude can generate incorrect or misleading responses and recommends checking the underlying sources.
That makes attribution especially important. Our webinar research found a strong pattern among highly cited content: lead with specific data, attribute the source, and then add interpretation.
For brands, that means avoiding unsupported “data says…” claims. Link back to original research wherever possible. Give dates and methodology. Correct outdated claims when you find them. Treat factual accuracy as part of GEO governance.
How should you benchmark your current AI visibility?
Do not start with a generic industry score, but with the questions that influence your own reputation and commercial outcomes.
Build a prompt set around areas such as category discovery, product comparisons, buyer questions, executive reputation, brand claims, competitors, customer pain points, and industry topics, then measure consistently across your priority AI platforms.
Useful metrics include:
- Citation share: How often are your brand and sources cited compared with alternatives?
- Domain rank: Where does your domain appear among sources used for important prompts?
- Prompt coverage: Which questions consistently include your brand, and where are you absent?
- Source mix: Which websites, social platforms, media outlets, review sites, and other sources influence the answer?
- Sentiment and narrative: How does the AI describe your company, products, executives, and competitors?
- Competitor visibility: Which competitors appear where you do not?
Those are also the kinds of signals Meltwater GenAI Lens is designed to track.
Engagement metrics still serve a purpose on LinkedIn, but they answer a different question. LinkedIn’s native analytics reports metrics such as impressions, reach, reactions, comments, reposts, saves, and article views. It does not currently present AI citation counts as a native post metric.
For owned websites, measurement is evolving quickly. Google recently introduced a dedicated Generative AI performance report in Search Console covering impressions from AI Overviews and AI Mode.
What does success look like for executive LinkedIn content?
There is no universal GEO benchmark for a C-suite LinkedIn program. Follower count and engagement rate can tell you how content performs socially, but GEO needs a second measurement layer.
You might track the number of cited URLs from the executive, citations per published piece, the percentage of priority prompts that surface their content, the subjects where they appear most often, citation growth over time, and which AI models use their work.
That also helps avoid chasing volume. Ten generic executive posts may achieve less than one strong article that answers an important industry question with original data and firsthand experience.
How should we think about reposting the same content from company and personal profiles?
Avoid simple duplication, because publishing exactly the same copy through multiple accounts creates little additional informational value. A better approach is to build several perspectives from the same underlying asset.
The company can publish the authoritative research. The CEO might explain the business implication. A product leader can discuss what practitioners should do about it. A customer-facing employee can answer the question they hear most often.
One insight becomes several genuinely different pieces of information. That approach also respects the central finding of our research: AI citation visibility comes from useful expertise, rather than from repeating the same marketing message as widely as possible.
What about comments and participation in trending conversations?
Comments are valuable for networking, visibility, relationship building, and understanding the questions people are asking. We did not, however, test comments as an AI citation format.
If your objective is GEO, turn the insight behind a strong comment into durable original content. A thoughtful comment that sparks a useful discussion could become a post. A recurring question in a trending conversation could become an article or FAQ.
Use conversations as research for content, rather than assuming comment activity itself will increase citation visibility.
How much should brands invest when AI source behavior changes so quickly?
This concern came up more than once during the webinar. Avoid building your entire GEO program around one model, one platform, or one content trick. Our own research showed meaningful differences between models, and citation behavior changed even during the research period. That volatility is a reason to build a flexible program.
Understand the questions your audience asks. Create accurate, useful answers. Develop credible expert voices. Make proprietary knowledge public where appropriate. Earn third-party validation. Measure how major AI platforms represent you. Review the data regularly.
Specific source rankings will change, but those practices are always going to be useful.
The same applies when headlines suggest that one AI platform is gaining or losing ground. Google AI search, ChatGPT, Claude, Copilot, and other products have different distribution models and use cases, so broad market-share comparisons do not necessarily tell you which platform influences your audience.
Measure the platforms relevant to your customers rather than choosing one winner.
Does this research apply to nonprofits or the public sector?
Our original research focused on 16 B2B categories. It was not designed to produce statistically robust conclusions specifically for nonprofits or public-sector organizations.
We therefore would not take the category-level results and present them as nonprofit or government benchmarks.
Build a sector-specific prompt set, identify the sources AI platforms use for those questions, then analyze which organizations, experts, media outlets, government sources, academic institutions, and other domains dominate the answers.
That is a much stronger approach than assuming patterns from commercial B2B categories will transfer unchanged.
What should brands do next?
The best place to start is relatively small - choose three to five topics where you need your organization to have authority. Identify the questions people ask about those topics. Benchmark the current AI answers. Look at the sources that already influence them.
Then identify the people inside your organization who can genuinely improve those answers.
Give those experts the research, data, and support they need to publish useful content. Use Company Pages and owned websites for authoritative brand information. Build content around real questions, comparisons, frameworks, evidence, and decisions.
Finally, measure the result.
GEO will continue to evolve, but that does not require brands to wait for the environment to stabilize. The useful question is simpler: when an AI platform answers an important question in your category, does your organization contribute something worth citing?
FAQ
Does paying to boost a LinkedIn post improve AI visibility?
There is currently no evidence from our research that boosting directly increases AI citation likelihood. Paid promotion can increase human reach and engagement, but AI citations appear to depend more strongly on the relevance, usefulness, structure, credibility, and accessibility of the underlying content.
Are LinkedIn articles better than regular posts for GEO?
Both can work. Text posts accounted for more citations overall in our research, while long-form articles generated more citations per piece. Use regular posts for consistent expert participation and articles for deeper, structured answers to important questions.
Should employees repost Company Page content?
They can, but adding substantive original commentary is more valuable than a simple reshare. Our research found original content represented 72% of cited LinkedIn content.
How often should brands measure AI visibility?
Monthly monitoring is a sensible starting cadence for most brands, with more frequent checks around launches, crises, major announcements, or fast-changing categories. Keep the core prompt set stable enough to identify trends over time.
What is the most important LinkedIn GEO tactic?
Publish genuinely useful expertise. Identify the questions your audience asks, choose the person best qualified to answer them, support the answer with specific evidence, and structure the content so both people and AI systems can understand it.