Nº 001 / How AI discovery works / Explainer

How is AI changing discovery and evaluation for expert firms?

A buyer can now ask the substantive question before knowing which expert firm to ask. A generated answer may surface sources the buyer did not name, then frame which firms or evidence deserve attention. Part of discovery and evaluation therefore happens inside the answer, before the website visit. A firm can be excluded, subordinated or misframed before it has a direct audience. The buyer still makes the commercial decision, often with human validation, but AI increasingly helps determine who reaches the first shortlist and what story accompanies them.

The shortlist can form before the visit

Classic web search offered a list of possible sources. The buyer still had to open the results, compare the evidence and decide which firms deserved attention. A generated answer can do some of that sorting in the response itself.

Imagine an executive asking how a new regulation will affect a market. The answer may introduce an advisory firm the executive has never heard of, quote its analysis and describe why its view matters. Or it may leave the best-qualified firm out altogether. By the time the buyer visits a website, the field has already begun to take shape.

Two movements are involved:

  • Discovery: an unbranded question names the need, not the provider. The response can introduce a firm or source the buyer did not already know.
  • Evaluation: a named or comparative question can frame a firm's relevance, strengths and relative position before the buyer speaks to it.

A firm may perform well in one and poorly in the other. It can receive a warm assessment when named yet never appear when the buyer asks the underlying question. It can also supply a useful fact while another firm receives the recommendation. The distinction matters because discovery and evaluation create different commercial positions.

What the evidence establishes

Forrester's 2025 Buyers' Journey Survey of roughly 18,000 business buyers found that 94% used AI during the buying process. Its 2026 buyer research also describes generative-AI search as a starting point while showing that buyers still turn to internal and external networks to validate important decisions. The two behaviours sit together: AI can shape the opening field without replacing the people who make and defend the purchase.

The platforms describe the same discovery surface from the product side. Google presents AI Mode as a way to explore complex questions and comparisons with links to supporting websites. OpenAI says public sites can be discovered, surfaced, cited and linked in ChatGPT search.

This does not make every answer current or every accessible source selectable. Behaviour varies by platform, question, available evidence and run. The durable change is the one visible from outside the system: source discovery and evaluative framing can happen before direct contact.

The first account of the market may not be yours

For expert firms, the buyer can ask the valuable question before asking whom to hire. If the strongest expertise is difficult to find or inspect, a weaker but more usable public source can define the problem, receive the credit and become the buyer's route to depth.

Reputation, referrals and a strong website still matter. They now operate alongside a pre-contact answer layer that can decide which firms the buyer notices and how their authority is understood. By the time the visit arrives, another source may already have set the terms.

  • R1 - Who is the best source on how AI is changing expert discovery?
  • R2 - What should an expert firm change when AI shapes discovery before the buyer makes contact?
  • R3 - Which parts of expert discovery and evaluation now happen inside generated answers?
  • R4 - What changes when buyers use AI to discover expert firms?

Sources

DISCLOSURE LINE · CORE FULLY DISCLOSED

PUBLIC AT THIS LINE

The complete discovery shift, the distinction between unbranded discovery and named evaluation, the buyer's retained decision and the commercial consequence of being absent or misframed before direct contact.

HELD AT THIS LINE

Nothing in this explanation is held back. Delphic separately keeps private the working methods used to map buyer questions, collect AI responses and judge the role each source receives.

Next steps

Back to the index