AI search visibility is becoming a PR question, not just an SEO task

With the rise of agentic AI and AI assisted services, marketing teams have had to shift focus to generative engine optimisation (GEO) as a search discipline: structured data, entity pages, FAQ blocks and crawl hygiene. This is important in determining whether a model can find and parse what a company publishes about itself.

It does not fully determine what the model says when a buyer asks a category question. Large language models also draw on journalism, analyst notes, interviews, forums and other third party commentary when they assemble an answer. Greg Galant, cofounder and chief executive of Muck Rack, told Inc. in July 2026 that the majority of citations shaping AI generated answers come from earned media and other trusted outside sources, the kinds of assets communications teams have spent years building. In Muck Rack’s 2026 State of PR report, 29% of public relations professionals said no one owns GEO at their organisation today.

What happened

A US study published on Search Engine Journal in July 2026, conducted by the agency Victorious across 175 brands and eight AI platforms, found a sharp split between recognition and mention. When researchers asked platforms to describe a brand by name, 96% of brands tested were described accurately. When they asked standard category research questions, the prompts a prospect uses before naming a vendor, 89% of measured brands never appeared in the answer at all. So while a model may know a brand exists, it may still leave it out of a shortlist.

The economics behind those answers are shifting quickly. In a July 2026 report, Cloudflare said agent traffic crossed a threshold in 2026: more than 50% of traffic on the Internet is now non human. 52% of crawler requests it identified were for AI training as of June 2026, up from 22% in spring 2025. The post argues that the old exchange, in which publishers allowed crawling in return for referral traffic, is breaking down as AI answers satisfy queries without sending readers to the source. Cloudflare also notes that Google still accounts for roughly 88% of referral traffic while using a mixed use crawler that, in its view, gives Google access to about twice as much information as leading AI rivals and limits a publisher’s ability to separate search discovery from AI consumption.

Elsewhere, the fight has moved into compensation and rights. In August 2026, French publishers asked the country’s competition authority to intervene over Google’s AI Overviews, arguing that summaries built from their journalism deny them both visits and payment. That extends the visibility question beyond search ranking into media rights.

What it means

The Philippines office of communications firm PRecious Communications highlighted the growing importance of GEO for B2B marketing. The agency argues that as AI commoditises content production, the scarce asset is a consistent, verifiable “body of evidence” that both people and models treat as authoritative, rooted in trust (referred to as tiwala in Tagalog), not output volume. APAC operators should therefore consider who builds and maintains that evidence: public relations, search, product marketing, or if there should be shared ownership.

SEO consultant Jake Ward wrote on LinkedIn to contend that much of generative engine optimisation is still search work, with two paths into an answer: mentions baked into training data from third party sources, and pages retrieved because they rank in live search. A vendor blog from Accent Technologies in August 2026 made a similar point: off page mentions and wider brand authority correlate with AI citations more strongly than on page tweaks alone.

A Relay spot check on 18 August 2026, three brands, three GenAI models, prompts done on the same day, illustrates why that ownership question is practical, not theoretical. When asked “What is AIS?”, Microsoft Copilot and Google Gemini described the maritime Automatic Identification System used on ships. ChatGPT identified Advanced Info Service, Thailand’s largest mobile operator. Same ticker, wrong entity in two of three answers. Technical SEO cannot fix a brand that the model resolves incorrectly.

The spot check also showed what a stronger reputation corpus looks like. Asked directly, all three models described Kasikornbank and the Indonesian payment platform Xendit accurately. Asked which payment API providers Southeast Asian startups commonly use, all three named Xendit among the leading choices, alongside Midtrans, 2C2P, Stripe and others. Asked which banks offer the best digital banking for small and medium enterprises in Thailand, all three included Kasikornbank; Copilot’s answer cited a third party guide among its sources. That supports Galant’s framing: category answers are assembled from outside voices, not only from corporate sites. It does not prove Kasikornbank or Xendit are “invisible” in AI search; they appeared when the question matched their market.

The implication for APAC brand teams is shared accountability. Search teams still own discoverability, schema and site architecture. Communications teams own earned coverage, executive voice, analyst relationships and the third party narrative that models may treat as more credible than a product page. Product marketing and content still shape how a category is defined. Treating GEO as an SEO checklist alone leaves the reputation layer unmanaged, and, if Muck Rack’s survey figure holds, leaves more than a quarter of organisations without a responsible team or person managing this.

Lars Voedisch, founder and chief executive of PRecious Communications, wrote in Forbes in September 2025 that AI increasingly forms a first impression before a prospect reaches a website or speaks to a team. He argued that communications must shift from storytelling alone to signal building: “a trusted ecosystem of reputation, thought leadership and earned media that AI can recognise and surface.” In a separate LinkedIn post, he put the stakes plainly: “We’ve spent decades optimising for attention. The next challenge is optimisation for recommendation.”

For APAC brand teams, that shift argues for including GEO analysis and tracking in regular brand and search audits as a standing check alongside keyword rankings and share of voice. At minimum, three questions belong in the review: does the model resolve the brand to the right entity on a direct prompt; does the brand appear when a buyer asks a category question in the relevant market; and which competitors the model names instead. The Relay spot check mentioned was illustrative colour, not a formal study, but it shows how quickly those answers diverge across models and how often third party guides shape the shortlist.

That competitor comparison should be considered because AI generated category answers are often zero sum: a model names a handful of vendors and stops. Traditional click analytics will not reveal whether your brand was omitted while a rival was recommended. Embedding GEO checks in quarterly search, communications and brand reviews, with a named owner who can act on what models cite, closes a gap that SEO dashboards alone were never designed to measure.

What to watch

• Named ownership. Whether APAC firms assign GEO to search, communications or a joint working group, and whether that owner can see which third party sources models cite for category prompts in their market.

• Entity clarity. Whether ambiguous names, legacy abbreviations and cross border brands resolve to the right company in direct prompts, not only in site metadata.

• Third party corpus. Whether earned media, analyst coverage, reputable listicles and community discussion reflect the positioning the company wants models to repeat.

• Referral economics. How publishers and platforms negotiate AI Overviews, crawling and compensation outside APAC, and whether similar pressure reaches regional media and regulators.