Computational Marketing
FeaturesLong read

What a 2,000-person marketing conference tells buyers about the Marketing Engineer and making your site legible to AI agents

A new role is emerging to make brands retrievable by AI buyers.

· 9 min read
Cover illustration for “What a 2,000-person marketing conference tells buyers about the Marketing Engineer and making your site legible to AI agents”
Features · November 13, 2026 · 9 min read · 1,957 words

A well-attended marketing conference circuit points to a change in who evaluates a brand before a purchase. When AI agents narrow shortlists, a persuasive website must also be accessible, specific, and easy for machines to retrieve. The emerging Marketing Engineer closes that gap by connecting content, technical infrastructure, human judgment, and measurement into one working system.

The conference circuit that declared AI agents a buying channel

London on Oct. 22, Sydney on Oct. 27, and Singapore were three stops built around one argument. The Zero Click World Tour treated AI systems as a separate audience to sell to. Its free, invitation-request, in-person events dealt only with AI search and buying journeys. Across the circuit, senior marketers attended from companies including The Estée Lauder Companies, Nestlé, Barclays, LinkedIn, Anthropic, Atlassian, Unilever, L'Oréal, and DHL Express. The range of companies made this more than a specialist search discussion.

The event’s stated mandate was to make brands legible to agents because agents were already transacting on behalf of customers. That is a different premise from using AI to write campaign copy. It places AI on the other side of the transaction, evaluating what a company offers. The conference positioned 2026 as the point when teams needed to instrument for this buying journey. Marketing buyers need to assess whether their current systems can serve an audience that retrieves evidence.

What "AI agents as buyers" actually means mechanically

An agent can narrow a product shortlist, and sometimes complete a purchase, before a person types a query into a search box. The underlying discovery mechanism is Retrieval-Augmented Generation, or RAG. In this process, an AI system pulls material from live pages before composing an answer. A brand enters consideration when its material is retrieved, used, and cited in that answer. The website therefore has a job before any visit from the eventual buyer.

At the moment of the query, a crawler or grounding system must be able to reach the content and extract something relevant. If that content supports the answer, the brand can become part of the buyer’s shortlist. Technical access and usable evidence now shape a commercial opportunity that may never begin with a conventional search click.

Gartner forecasts that one in five digital commerce transactions will be executed through AI platforms or agents by 2030. It gives the conference a reason to press teams to act now, while allowing for buying journeys to change at different speeds.

Why ranking on Google and getting cited by an AI answer engine are different jobs

Being visible in traditional search still helps a page become a possible source. Accessibility and search rank are important signals for AI citation, so the new task does not replace SEO. But retrieval introduces a second test. AI systems pull out chunks that contain enough information to answer a question on their own. A page built around keyword density or a long narrative may rank yet give a grounding system no passage it can quote as an answer.

A nosnippet meta directive can suppress AI citation because engines rely on snippet data during grounding. It is a configuration choice that a content team may never see, so improving copy alone cannot solve the problem. Search performance and answer inclusion overlap, but they are not interchangeable outcomes.

The temptation is to rewrite everything for the summarizer. Luciano Viterale, Head of Growth APAC at Rippling, addressed the concern in a Zero Click Sydney session titled “Don't Tank Your Site Chasing ChatGPT”. Content written to game an AI summarizer risks serving neither people nor machines well. The more useful distinction is between generic claims a model can reproduce and original evidence it has a reason to retrieve. Survey results, proprietary case studies, and named expert opinions from real conversations give a source that distinction.

GEO, AEO, and AI visibility as layers of the same infrastructure, not competing tactics

AI visibility is the outcome in which a brand appears in AI-generated answers. GEO is the content and structural practice that makes that possible. AEO fine-tunes answers to specific questions within that structure. SEO helps systems find the brand, while AI-brand reputation, sometimes called AIO, deals with how they understand and describe it.

The distinction has practical consequences for how teams work. Treating GEO as a separate content track can lead to AI-specific variants published beside regular content, without improving either. That approach risks scaled-content abuse flags from search engines as well as weak experiences for readers. Each discipline addresses a different part of how the same brand gets found.

Coordination is harder than agreeing on the vocabulary. PR shapes external accounts of the brand, content supplies explanations and evidence, SEO handles discoverability, and product marketing defines what the offering means to buyers. GEO forces those responsibilities to converge because AI answers draw on the understanding they collectively create. Who owns the gaps between them? That unresolved organizational question helps explain the emergence of the Marketing Engineer.

The Marketing Engineer: what the role is and why it emerged from this specific pressure

The Zero Click conference gave this definition of the role. “A Marketing Engineer is part builder and part artist — the artist understands the brand, the audience, the product; the builder connects the right data and tools, designs the workflow around the problem, and decides when a human needs to make a decision. Marketing Engineers are marketers first.”

The final sentence anchors the whole definition. Technical ability serves an understanding of the audience, product, and brand. The role differs from the earlier marketing technologist, whose work was mainly to set up platforms and keep integrations working. A Marketing Engineer designs the engagement between people and AI, including the points where automation must yield to judgment. Configuring software is only part of that work.

The pressure comes from a change in marketing itself. Individual campaigns remain outputs, but the larger task is designing adaptive systems that can learn, optimize, and execute with human oversight. Fewer job listings for content marketing manager roles hint at where the job market is heading, though they do not prove that one title directly replaces another. The stronger explanation lies in the work. When discovery depends on editorial evidence, technical access, external reputation, and repeated testing, someone must design how those pieces operate together.

What the Marketing Engineer actually builds: the pipeline as primary artifact

The Marketing Engineer’s primary artifact is a pipeline. The role builds and maintains the agents and automations that produce, distribute, and measure marketing work. It also sets rules for brand voice and reviews output with a marketer’s eye before customers encounter it. The person designing the system remains accountable for the quality of what it produces.

The basic architecture runs from brief to generation, validation, and publication, on a schedule with defined review checkpoints. Agents can draft, design, and schedule, while a person approves work before it reaches a customer. Everything an agent produces goes through a quality check before it is put to use. High-stakes actions, including budget changes and audience suppression, also need approval before they go ahead.

That design produces more than publishable work. The pipeline creates a traceable record of what was generated, approved, changed, and published. This audit trail makes the workflow a measurement instrument as well as a production system. The EU AI Act, NIST guidance, and FTC scrutiny of AI-generated claims form part of the governance context, though they are not interchangeable rules. Traceability gives the team a basis for examining how an output reached a customer.

Why the human review gate is non-negotiable, not a performance concession

Automation without quality control can produce generic, inaccurate, or misattributed output. A claim can read well and still be wrong. AI-generated drafts can contain hallucinations. The system cannot treat completion as validation. A review gate separates work that has merely been produced from work the brand is prepared to stand behind.

But does review give up the speed automation promises? Not when the pipeline assigns drafting to AI and judgment and approval to people. This architecture avoids asking a slow agency process to meet the publishing cadence of AI discovery, while also avoiding the quality risk of unreviewed output. Human review has a direct link to GEO because AI answer engines weigh how authoritative and trustworthy a source is when choosing what to cite.

A factual article can still leave unanswered questions about its origin, approval, or the responsibility for its claims. Provenance and auditability address those questions, while editorial review addresses what the customer will read. Neither substitutes for the other. As oversight requirements mature, a pipeline without human approval leaves a governance gap even when an individual output happens to be correct.

Making your site machine-legible: what the ground-site strategy requires structurally

A “ground site” is an owned web presence structured so AI grounding systems can retrieve, verify, and cite it with confidence. The term describes a role. Its purpose is to become a usable source when an AI needs evidence for a claim about the brand’s category. For a team making changes, technical access comes first. URL accessibility, crawlability, and snippet-preview settings need attention before more content is commissioned, because inaccessible material cannot do its job.

Once the system can reach a page, it needs a complete answer to extract. The practical response is to write answer-sized sections under clear headings, with enough substance for each section to stand on its own. Definitions, bullet points, and summaries can help both readers and chunking systems understand the material. This is good information design rather than a separate machine-only writing style, and it need not sacrifice the page’s larger argument.

Structure alone does not give a source distinction. Original survey results, proprietary case studies, and opinions from named experts in documented conversations give a source evidence that generic generated text does not have. Build that evidence into the accessible, answer-sized sections to earn citations. A model needs something worth attributing to the source. More text is not the same as more evidence.

The owned site is also only part of the grounding environment. Web mentions correlate with AI visibility, making unlinked brand presence relevant as well as backlinks. Press coverage, podcast appearances, forum discussions, and third-party reviews contribute external signals of authority. Distribution therefore belongs in the site strategy. It helps connect the brand’s own account of itself with the wider information environment AI systems use.

Measuring whether any of this is working across the engines that matter

Publication does not establish AI visibility, and stable search rankings do not establish it either. The relevant test is whether ChatGPT, Claude, Gemini, and Perplexity name and cite the brand in response to category queries. Engine behavior varies substantially, so the gap between presence in one and presence in another requires assessment beyond analytics alone. The Marketing Engineer’s responsibility therefore extends beyond running the publishing pipeline to testing its discovery outcomes.

In practice, this requires a standing set of category, comparison, and use-case prompts run across each engine on a schedule. Query monitoring records whether the brand is named, cited, or absent. Citation tracing then checks whether a named source actually points to owned content. Hallucinated or misattributed citations need to be identified before a team can address them. Trend tracking adds the time dimension, distinguishing a recurring pattern from a single answer.

After a new piece of original evidence is published, subsequent runs can show whether the citation signal moved. A single change does not prove its cause. Repeated measurement nevertheless gives the team a firmer basis for improving the system than publication counts or assumptions about what engines prefer. That is the final test of the Marketing Engineer’s work, whether the system helps a brand become a retrievable, trustworthy source in the answers that shape consideration.

More in Features