AI has become the single most important decision-shaping layer in the startup world. ChatGPT, Claude and Gemini are now the default routes through which prospective clients size up your products and applicants consider your job postings. Your potential investors are firing off prompts rather than spend time scrolling through several
Julija Jegorova
AI has become the single most important decision-shaping layer in the startup world. ChatGPT, Claude and Gemini are now the default routes through which prospective clients size up your products and applicants consider your job postings.
Your potential investors are firing off prompts rather than spend time scrolling through several pages of Google search results. AI compresses discovery, evaluation and recommendation into a single interface, and navigating this new reality will prove decisive for startups.
What LLMs are reading
To understand how to make a startup appear in LLM results in the best possible light and as relevant answers to prompts, we need to understand how LLMs work.
In a May 2026 report, the PR analytics firm Muck Rack found earned and third-party media make up about 84% of AI citations, with paid and advertorial content comprising a rounding error of 0.3%. Ahrefs, an SEO data company, studied 75,000 brands and reached the same conclusion. A March 2026 distribution study by MarTech company Stacker found that pushing the same content through earned outlets rather than your own site lifted AI visibility by a median 239%.
In conclusion: earned, corroborated coverage is now the start of a supply chain of both trust and visibility within LLMs.
But many founders make the mistake of pursuing the big-name publications straight away. Seasoned PR professionals can tell you that this is rarely a feasible strategy, and the studies’ investigations into earned media incidence serve to confirm the importance of an approach that was already more efficient.
Trade and industry media, longtime quiet anchors of trust and continually relevant, are actually within grasp for early stage startups and rewarded by LLMs for their subject-specific authority and depth.
In addition, recency always plays a role in how results show up. Smaller, trustworthy publications offer the opportunity of a more consistent, steady storytelling journey.
What you can and cannot control
Community and reference platforms among the most-cited sources of all, and they’re un-pitcheable. Muck Rack found that Wikipedia appears in ChatGPT’s top three cited domains in 12 of 17 industries, and Reddit is cited heavily across models. Both are notoriously hard to influence. Wikipedia vets editors for impartiality and requires high-trust, earned or academic citations.
Review platforms pull serious weight too. A Seer Interactive analysis of 800,000 AI responses found review sites among the most-cited sources for product questions, with G2 ranking near the top for B2B software.
However strategically you maximise these channels, AI is nevertheless gravitating towards something less gameable. Building domain authority with backlinks was not inherited, although it still works for SEO. The Ahrefs study found brand mentions predicted AI visibility three times more strongly than backlinks.
You can’t buy your way into AI citations either. AI doesn’t reward any paid media to reach the demographic of your choice. Its sources are harder to engineer.
One thing you can control is owned media, such as your website, blog and social media. It accounts for 16% of citations and is used by models to fill in any informational blanks.
Appropriate optimisation includes the hygienic foundation of a crawlable site and a tidy social presence. AI aside, many of these elements are indispensable to making a good first impression on a journalist. To better cater to AI models, find out how to implement structured data and schema markup, headings and question-and-answer formats.
Thinking bigger about comms and AI
LLMs are doubtlessly adept at synthesising and regurgitating information (and increasingly also solving problems), but they struggle to perceive and absorb reliable information about the real world. There are multiple unfortunate instances of models learning from their own error-riddled AI outputs and creating a doom loop of slop. Chronicling reality is still a human job, as is decision-making.
Comms is not just about AI. The necessary work of adapting to this new paradigm shouldn’t detract from regular PR work (which boosts the quality of AI citations) or create tunnel vision around the other forms of trust-building where AI is less relevant.
A growing layer of credible human voices increasingly sets the tone for trust in the B2B space. These are subject-matter experts, operators and practitioners, analysts, newsletter writers and podcasters with engaged, expert audiences, in and out of traditional media. They prove that domain-specific authority trumps audience size and sponsored reach.
Those audiences don’t just exist in the ether. They attend events and interact with people in real life. That level of in-person, spoken, real-world engagement is the comms ace-in-the-hole. A founder on a panel or a stage is a leader with a point of view, in front of and trusted by the exact people who matter. As AI grows in importance, don’t underestimate IRL.
What changes, what stays the same
We’re told writing for machines is now as imperative as writing for humans. That’s a misrepresentation. Only after you win the humans behind the trusted platforms will the machines follow.
Model preferences aren’t fixed, and the deals between publishers and AI companies are still being negotiated. The introduction of ads to LLMs will create new rules. Change is coming, and it will keep coming.
What is less likely to change is the basics of modern trust-building through good comms, which also happens to increase your AI visibility. In the end, the platforms and publications that AI sources the most are those people already trust, and earning that trust has always been the heart of comms work.
It bears mentioning that the fastest way to undo all of the above is to use AI badly in your outreach: AI-generated slop pitches inevitably undercut the very trust you are trying to establish, and journalists and audiences punish it quickly. As City AM editor-in-chief Christian May recently wrote on PR Week: “We will never be a dumping ground for your AI slop.”
Smart startups will heed that gentle (but firm) warning. But more how to PR within your own workflows in comms, for another time.
