Marketing to the Agent Gatekeeper
Make marketing useful to machines so agents can recommend it to humans.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 98%
The agent-gatekeeper framework treats AI systems as an additional audience between a company and its customer. Agents may prioritize inboxes, summarize brands, inspect websites, evaluate trials, and recommend a shortlist before a human engages. Marketing must therefore provide accurate, accessible, and machine-readable evidence about capabilities, limitations, pricing, and fit. The process begins by mapping where agents mediate attention, then making those touchpoints easy to retrieve and evaluate. Marketers test how models interpret the material and correct ambiguity or unsupported claims. Human-oriented creativity remains essential, but it operates after or alongside an agent-facing information layer designed to survive filtering, comparison, and summarization.
Origin
Extracted from Marketing Against The Grain while the hosts analyzed AI-prioritized email and agents evaluating brands on behalf of customers.
Core principles
- 01Agents increasingly filter what humans see.
- 02Machine comprehension becomes part of audience targeting.
- 03Clear factual marketing survives summarization better than empty persuasion.
- 04Human creativity still matters where people remain in the loop.
- 05Agent-facing structure and human-facing appeal must work together.
How to run it
- 1
Map agent-mediated touchpoints
Identify where an assistant may filter, summarize, compare, or act on marketing before a person sees it.
Pro tip Start with inboxes, search, websites, product documentation, and trials.
Watch out Do not assume the agent only reads conventional landing pages.
- 2
Define the decision evidence
List the facts an agent needs to determine fit, including capabilities, constraints, pricing, proof, and exclusions.
Pro tip Answer comparison questions directly and consistently.
Watch out Hidden limitations can produce rejection or loss of trust.
- 3
Make evidence digestible
Present information in clear language and structures that can be reliably retrieved and summarized.
Pro tip Use descriptive headings, explicit statements, and consistent terminology.
Watch out Keyword stuffing can reduce clarity for both machines and people.
- 4
Simulate agent evaluation
Ask representative models to summarize, compare, and recommend the offering from the available materials.
Pro tip Test with both ideal-fit and poor-fit customer scenarios.
Watch out One model's response is not proof of universal interpretation.
- 5
Improve the human handoff
Ensure that people who pass through the agent's filter encounter persuasive, creative, and trustworthy experiences.
Pro tip Align the human experience with the evidence the agent already presented.
Watch out Agent optimization cannot rescue a disappointing product or misleading message.
In the wild
Apple Intelligence evaluates incoming email importance. A vague promotional message falls below urgent and useful correspondence, while a concise message tied to the recipient's needs has a better chance of being surfaced.
→ Relevance and information quality affect whether the human sees the campaign at all.
A buyer asks an assistant to investigate several products. The agent reads websites, analyzes trials, summarizes strengths and limitations, and recommends a shortlist. Vendors with clear factual material are easier to evaluate than those relying on vague claims.
→ Machine-readable evidence becomes part of demand generation and competitive positioning.
Common mistakes
Optimizing only for human clicks
An agent may decide whether the human encounters the message before click-through tactics can work.
Publishing vague superlatives
Unsupported claims provide little usable evidence for an agent performing comparisons.
Forgetting the human experience
Machine readability earns consideration, but people still need trust, emotion, and a strong product experience.
Is it for you?
Best for
It is best for email, B2B, product-led, search, and comparison-driven marketing exposed to AI mediation.
Not ideal for
It is not ideal as a replacement for emotional brand building or direct human relationships.
From the transcript
“AI is probably gonna be the best filter for bad marketing that has ever existed.”
“how do you construct or funnel that it's easy for an AI agent to digest and find what it's looking for?”
“We are going to be marketing just as much, if not more, to agents and AI as we are to humans.”
From the episode
Apple Intelligence Can Outsmart 99% Of Marketers (WWDC Recap)