A BCG analysis compared the results of traditional search engines with those of AI-powered answer engines. The overlap is between 8 and 12%. A company can appear at the top of Google and be entirely absent from ChatGPT, Perplexity, or Copilot responses.
The figure becomes more significant when you look at what produces it. BCG found that 97.2% of citations generated by AI engines cannot be explained by backlink profiles, the currency on which online visibility has been built for the past twenty years1. It's not that AI systems penalize companies that rank well on Google. It's that they use different criteria to decide who to cite, who to recommend, who to surface in a response. And the central criterion is readability: structured data, verifiable information, content that a machine can interpret without ambiguity.
For businesses, this changes the rules of visibility. And not just in digital marketing.
When an AI agent evaluates a supplier, a product, or a service, it doesn't browse a website the way a person would. It doesn't read brochures, isn't impressed by design, has no prior relationships or brand loyalty. It queries data. If the data is structured, it compares. If it isn't, it moves on.
PwC has formalized the requirements a company must meet to be selected by AI agents: discoverable (not just indexed, but interpretable), trustworthy (with accurate and verifiable product data), and structured (with processes on which an agent can operate reliably)2. This isn't a theoretical framework: PwC and Stripe built an operational infrastructure for agentic commerce together, starting from exactly these three requirements. Ian Kahn of PwC puts it this way: "AI agents are becoming the new storefront."
Catalist, in a study on B2B distribution, is more direct: companies that provide verified certifications, complete technical datasheets, and supply chain documentation in structured formats become preferred suppliers for procurement agents. Those that rely on PDFs sent by email and verbal reassurances get filtered out of the process3. It's not about product quality. It's about readability: if the agent can't verify, it won't select.
McKinsey, in its own research on agentic commerce, puts it in terms that leave little room for interpretation: if a company's catalog, policies, and value proposition are not machine-readable, agents (and by extension customers) will not find it, regardless of brand strength4.
This isn't just a B2B issue. A BCG/Moloco survey found that 33% of American adults discover brands through personal AI agents, and 47% already use AI tools to evaluate purchases1. PwC estimates that over a third of traffic generated by generative AI is replacing traditional search2. The filter deciding who gets seen and who doesn't is shifting, and it's shifting toward machines.
The parallel with the early days of SEO now comes naturally. But there's a difference that raises the stakes considerably.
In traditional search, poor ranking meant appearing on page two or three. Few clicks, but at least a presence. With AI systems, page two doesn't exist. Language models cite 2–7 sources per response. Those who don't make the cut aren't just ranked lower, they cease to exist entirely.
Gartner predicts that organic search traffic could be halved by 20285. Gartner itself, in a February 2025 analysis, estimated that by 2026, 60% of AI projects lacking adequate data will be abandoned5. This figure doesn't only concern those building AI. It also concerns those who need to be seen by AI: only 4% of surveyed organizations said their data is truly ready.
What does "AI-ready data" actually mean? Not volume. Structure: consistent definitions, complete metadata, machine-readable formats, verifiable information. Those who invested in SEO twenty years ago built an advantage that in many cases still holds today. Those investing in readability for AI systems are making the same bet, with a window that is probably narrower.
The deepest shift isn't technological. It's in how companies are evaluated.
Traditional marketing is built to persuade people. It tells stories, creates emotional associations, builds trust through repetition and recognition. AI agents don't work that way. Persuasion works on human operators, while AI agents are focused on comparison. They look for structured attributes, verifiable specifications, data that can be set side by side with a competitor's and processed objectively.
McKinsey sums it up with a sharp formula: verifiable data beats marketing gloss4. Products that are emotionally readable to people but semantically opaque to machines risk disappearing from agent-mediated flows. The difference is concrete: "superior quality, innovative design" says nothing to an AI agent. "6061 aluminum alloy, 2.4mm wall thickness, static load certified at 500kg" says everything.
This doesn't mean brand no longer matters. It means brand alone is no longer enough to guarantee visibility. Before convincing a customer, you have to pass a filter that cannot be convinced. And that filter reads data, not stories.
PYMNTS and Visa, in a report on the AI agent economy, describe the new reality well: every company now has two customers to serve simultaneously. One is the person, who moves on emotion and trust. The other is the agent, which moves on clean data, clear rules, and verifiable performance.
That 8–12% overlap between traditional search and AI responses mentioned at the opening of this article is not a flaw in immature systems. It's a signal that the terrain on which visibility is contested is shifting. For twenty years, "being findable" meant investing in SEO. Today it means making your data structured, readable, and verifiable for systems that don't scroll pages, can't be persuaded, and have no memory of past relationships.
Companies doing this aren't running an IT project. They're building the foundations of their relevance in a market where, for the first time, being good isn't enough. You also have to be readable.
Sources
[1] Boston Consulting Group, The Future of Discoverability (January 2026); BCG / Moloco, AI-Powered Brand Discovery Survey (2025)
[2] PwC, Agentic Commerce: Compete in an AI-Buyer World (October 2025)
[3] Catalist Group, The State of AI in B2B Distribution (2026)
[4] McKinsey & Company, The Automation Curve in Agentic Commerce (January 2026)
[5] Gartner, Lack of AI-Ready Data Puts AI Projects at Risk (February 2025); Gartner, Top Strategic Predictions for 2026 and Beyond (October 2025)