April 13, 2026

GEO Is Rewriting the Rules of Digital Visibility

Christien Paul, Espace M – ICOM’s AI Group

Members of the Global Artificial Intelligence Working Group within the international ICOM Network explore the evolution of AI in the industry.

For two decades, SEO taught brands to compete for rankings. Now, generative interfaces are teaching brands to compete for inclusion inside answers. In a recent AIcom session, Christien Paul, SEO Team Lead at Espace M, one of ICOM’s Canadian member agencies, argued that this shift is not a tactical tweak — it is a structural redefinition of what “being findable” means. GEO (Generative Engine Optimization) is the discipline emerging to meet that reality. And for independent agencies, it may be the next arena where expertise compounds.

From Rankings to Inclusion: Why GEO Exists

Christien’s framing was simple and unsettling: many marketing teams still assume that if a site is “good for Google,” it will be “good for AI.” That assumption is increasingly dangerous. As AI Overviews, chat-based discovery, and LLM-powered assistants become the first touchpoint for questions that used to start with a search engine, visibility is migrating upward — from the results page into the generated response itself.

In that environment, the competitive unit is no longer the blue link. It is the excerpt, the citation, the summarised answer — the part the model is willing to repeat. GEO, in Christien’s view, is therefore not a replacement for SEO but a new layer: a blend of technical accessibility, structured meaning, and content designed to be quoted safely.

The Invisible Visibility Killer: Security Defaults That Lock AI Out

number of websites are quietly blocking LLM access through security and compliance configurations, especially on Cloudflare, but also through firewalls, SSL setups, and bot-management policies.

The pattern is familiar in enterprise environments: protect the site, reduce bot traffic, allow the “known” crawlers — and accidentally treat LLM crawlers as suspicious unknowns. The business consequence is severe: you can invest in content, brand storytelling, and classic SEO, while the generative layer never even sees the material you’re publishing.

Christien’s GEO first checks were operational, not creative:

Cloudflare bot-management settings (are LLM crawlers blocked by default?)
Firewall rules that implicitly treat non-search-engine bots as hostile
SSL and security configurations that can break or limit access in unexpected ways
Robots directives that may be overly restrictive for the new discovery landscape

His emphasis was not to “open the doors to everything,” but to make access a conscious choice. GEO begins by ensuring the brand’s most valuable content is reachable by the systems now shaping discovery.

Schema as Strategy: Making Meaning Machine-Readable

Once access is possible, the next challenge is comprehension. Christien pointed to a quiet mismatch: modern websites are built for human experience — JavaScript-heavy interactions, accordions, dynamic rendering — while AI systems are not always consistent at retrieving and interpreting what’s hidden behind those layers.

In that context, structured data stops being “SEO hygiene” and becomes strategic infrastructure. Schema markup provides a clean, explicit map of meaning: what the page is, what questions it answers, what products or services it describes, and what entities matter. For GEO, schema functions as a credibility and clarity layer — reducing ambiguity so an LLM can cite with confidence.

Christien highlighted schema types that tend to pay off in generative environments: FAQ-style structures (used sparingly and intentionally), product and service schema where applicable, and local business and organisational signals.

He also noted a practical reality: many WordPress implementations are schema-light. They may look fine in traditional SEO audits but remain under-specified for AI extraction. His recommended implementation route for WordPress was SEOPress Pro, largely because it supports richer schema configurations beyond the bare minimum.

llms.txt: The Early-Adopter Front Door

Christien introduced llms.txt as an emerging practice — not yet standardised, but increasingly discussed — aimed at giving LLMs a cleaner path to a site’s most relevant content. His stance was pragmatic: when standards are still fluid, early adopters have outsized influence. Waiting for maturity can mean arriving after competitors have already shaped how their brands are represented.

In business terms, llms.txt is a low-cost hedge: it may become a common convention for AI-readable publishing, and the upside of being prepared is likely larger than the downside of experimenting early — provided it’s done responsibly.

The Compliance Trap: Cookie Walls That Hijack the Page

One of Christien’s most actionable points had little to do with marketing and everything to do with compliance UX. Some cookie banners and privacy modals embed long blocks of legal text directly inside the overlay. That design choice can inadvertently make the legal copy the most prominent, most extractable content on the page — burying the brand’s actual message beneath it.

His recommendation was straightforward: where possible, link out to privacy and cookie policies rather than embedding entire documents inside modals. In a GEO world, legal implementation patterns can directly affect discoverability and brand comprehension.

Designing for Quotability: Mini-Answers Beat Mega-FAQs

Christien’s content advice called for a shift in mindset. Historically, many brands treated FAQs as content dumping grounds — long lists of questions that exist mainly because SEO playbooks said they should. GEO changes the incentive: generative systems favour short, self-contained explanations that can be quoted safely.

His preferred pattern was the “mini-answer”: a crisp question (or question-like heading) paired with a brief, quotable answer placed directly inside the relevant service page. Not hidden in an FAQ warehouse. Not wrapped in marketing fluff. Designed to be extracted.

Why this matters commercially: it increases the likelihood of being pulled into AI Overviews and chat responses; it reduces misinterpretation by making the answer explicit and bounded; and it supports consistency — the brand’s official phrasing becomes easier for systems to reuse.

Hallucinations and Brand Risk: Visibility Requires Monitoring

The discussion turned to risk when participants raised a concern: if LLMs talk about you, do they reveal sensitive information — or worse, invent it? Christien’s response landed in the middle. Much of what appears in generative summaries (especially pricing ranges) often comes from public signals like reviews and third-party commentary rather than proprietary data.

But hallucination is real. He cited a practical example from a local-services context where an LLM incorrectly attributed a service to his client — a mistake likely driven by confusion with competitors. The implication for brands is clear: GEO is not only about getting included; it’s also about ensuring the story that appears is accurate. Christien referenced ZipTie as part of his toolkit for auditing how a brand shows up across generative interfaces.

Weaponised PR and Black Hat GEO: The Cycle Is Starting Again

The most future-facing portion of the session explored manipulation. If generative models summarise what they ingest, a predictable question follows: will brands try to “feed” the models narratives designed to be repeated? The group discussed early signs of precisely that — syndicated PR and content engineered less for humans and more for model ingestion.

A key nuance emerged: PR that mimics reviews can pollute comparative answers unless analysis is constrained to vetted sources. This is the generative-era version of the old SEO problem — low-quality signal flooding the system. Tactics discussed included syndicated PR campaigns designed to create quotable narratives for LLMs, opaque sponsored placements presented as neutral commentary, and the role of publication velocity (sudden spikes can look suspicious; steady, consistent activity tends to appear more credible).

Christien’s caution was clear: this is a high-risk path. The industry has seen this movie before. Black-hat experimentation tends to trigger enforcement and reputation damage. GEO will likely develop its own norms, penalties, and trust signals.

A Practical GEO Playbook: What Agencies Can Do in 30 Days

For independent agencies, the immediate opportunity is to run a cross-functional GEO audit — not as a marketing project, but as an operational initiative that touches IT, legal, and content teams.

A 30-day starter plan:

Run an access audit: confirm whether LLM crawlers can reach key pages (Cloudflare/firewall/robots/SSL).
Map and enhance schema: focus on core service pages, products, and organisational entities.
Pilot llms.txt responsibly: provide clean pathways to high-value content and references.
Refactor one service page using mini-answers: create short, quotable, precise Q&A blocks.
Set up monitoring: track how the brand is described in generative interfaces and flag inaccuracies quickly.

The Strategic Bottom Line

Christien’s closing message was not “do GEO when you have time.” It was: audit now, because the biggest risks are invisible until you look. In a world where customers increasingly begin their journeys inside generated answers, visibility becomes a multi-disciplinary asset — and a multi-disciplinary vulnerability.

The agencies that win in this environment will be those who can bridge security, structure, and storytelling — and who can keep the brand’s narrative accurate as machines begin to speak on its behalf. That is precisely the kind of expertise that compounds inside a network like ICOM.