Anshuman is a marketing director at Cognizant. Influences global audiences on AI strategy, enterprise transformation, identity and society.
getty
When a prospective client’s procurement team begins evaluating vendors, they no longer start with Google. They open ChatGPT, Perplexity or Microsoft Copilot and ask a single question: Who should we shortlist? By the time your sales team learns about the opportunity, the AI has already formed a view of your company. That view was assembled from sources you don’t control, without your participation or knowledge.
This is not theoretical. Over 1 billion people use ChatGPT monthly. Research from 6sense found that 94% of B2B buyers deployed generative AI during their most recent purchase cycle. For most enterprise brands, the AI-mediated first impression now precedes every other touchpoint in the buyer journey. And most organizations have built no framework for managing it.
Large language models form probabilistic understanding by synthesizing patterns across enormous volumes of text: news articles, analyst reports, employee reviews on Glassdoor and LinkedIn, Reddit discussions, product comparisons, regulatory filings, press releases and competitor claims about you in their own content. The synthesis produces something functionally equivalent to reputation: a weighted set of associations, which competitors it mentions alongside you, what capabilities it assigns you and how often it recommends you.
This reputation operates across three measurable dimensions. Accuracy: whether what AI says about you is factually correct. Depth: how much the model knows and how confidently it speaks. Sentiment: whether the characterization carries positive, neutral, or negative associations.
This is a governance problem, not a marketing problem, because of the mechanism. A 2026 analysis of AI brand representation found that 84% of AI citations derive from earned media sources outside your direct control rather than brand-owned pages. The majority comes from what others have said, published, reviewed and discussed in forums you do not moderate, publications you did not brief and contexts you did not anticipate.
The implications are material. When an AI model misrepresents your market positioning, misattributes client relationships or overstates capabilities, that error propagates across every buyer interaction the model participates in until retraining occurs. Unlike a newspaper correction or a social media clarification, there is no mechanism for real-time revision. The inaccuracy is embedded. For brands in regulated industries, the exposure becomes compliance risk. AI models have been documented misrepresenting pricing, certifications and service capabilities for financial services and healthcare firms, generating descriptions that would constitute misleading advertising if the company had published them directly. But the buyer read it, believed it and acted on it.
The marketing industry’s response has been rapid. Generative engine optimization (GEO) emerged as a discipline of structuring content so AI systems are more likely to cite it. Share of model voice emerged as a metric for tracking citation frequency across platforms. Both address the outbound half of the problem.
But they do not address the inbound half: what AI has already absorbed from the information ecosystem surrounding you—the ecosystem you share with critics, competitors, former employees, industry analysts and broader public discourse. Meltwater’s May 2026 analysis of 9.5 million AI citations across 16 B2B categories found that platforms brands do not control (LinkedIn, Reddit, YouTube) account for 47.5% of all citations. Company-owned content accounts for 18.7%. That ratio is instructive: The majority of what AI knows about your brand was written by people not on your payroll, on platforms your communications team does not govern.
This creates a specific category of brand risk. An organization can publish perfectly structured, authority-anchored content and still watch its AI reputation shaped by 3-year-old Glassdoor reviews, a critical analyst note or a competitor’s comparison page that positions it unfavorably. GEO optimizes what AI finds going forward. It cannot rewrite what AI already believes.
Machine Relations Operates On Four Fronts
• The first is monitoring: continuous querying of major AI platforms to understand current brand representation in terms of accuracy, depth, sentiment and recommendation frequency. This is not a monthly report. It is continuous intelligence, equivalent to brand-tracking studies.
• The second is source management: identifying which external sources are most heavily weighted in AI citation patterns for your category and deliberately cultivating relationships with those sources. A Muck Rack analysis found that the overlap between journalists most pitched by PR teams and those most cited by AI engines is just 2%. Most organizations invest their media relations budget in relationships that AI does not particularly value, while the sources that shape AI’s understanding remain unmanaged.
• The third is entity consistency: ensuring your organization’s name, capabilities, leadership and positioning are described consistently and accurately across Wikipedia, LinkedIn, industry databases, analyst registries and professional associations. Research confirms that inconsistent entity signals create entity confusion in AI models.
• The fourth is remediation: when AI misrepresents your brand, having a structured protocol for introducing corrective information into the information ecosystem in ways that influence subsequent model updates. This requires publishing accurate, authoritative, structured information in the sources AI weights most heavily, not simply issuing press releases.
The front that surprised me most was the simplest one. Monitoring sounds like instrumentation, something you set up once and review quarterly. In practice it works as a mirror, and clients rarely recognize the reflection. On one engagement, we ran the shortlisting query their buyers were already asking. The model described the company as [X]. Its entire strategic push for the previous three years had been [Y]. Nothing the model said was false. The emphasis simply belonged to someone else, and when we traced it, the weight was coming from [old coverage/a competitor’s comparison page/an analyst note].
That changed how I open engagements. I used to start with a content audit. Now I start by running the queries a buyer would run and putting the unedited output in front of the leadership team before anyone discusses strategy. Machine relations sits at the intersection of marketing, communications, legal and technology, reporting to the C-suite. The closest analogy is investor relations, a function responsible for managing how a specific and consequential audience understands and values the organization. Expect machine relations to be standard at every enterprise brand operating in a competitive information environment.
Forbes Communications Council is an invitation-only community for executives in successful public relations, media strategy, creative and advertising agencies. Do I qualify?
