Indonesia Singapore ไทย Pilipinas Việt Nam Malaysia မြန်မာ ລາວ
← Back to Blog

Google Gemini Is Winning the AI Referral Race — Now What?

Optimise for Gemini's entity-driven citation model now — brands visible in structured, authoritative content will capture AI referral traffic as it scales.

Editorial illustration of a grizzly bear quietly rearranging search result cards on a wall while other figures sleep
Illustrated by Mikael Venne

Google Gemini is outpacing Perplexity and closing on ChatGPT for referral traffic. Here's what that means for your GEO strategy in Southeast Asia.

Google Gemini more than doubled its referral traffic to websites in just two months. If your GEO strategy is still centred on ChatGPT, you may be optimising for yesterday’s leaderboard.

Gemini’s Referral Surge Changes the Citation Calculus

SE Ranking’s data, reported by Search Engine Journal, shows Google Gemini has overtaken Perplexity as a referral traffic source — and is closing the gap on ChatGPT, which has declined from its peak. This isn’t a minor footnote. It’s a structural signal about where generative search authority is consolidating.

Gemini’s referral architecture differs meaningfully from ChatGPT’s. Because it sits inside Google’s ecosystem, it draws heavily on the entity graph, structured data, and the same quality signals that power traditional Search Quality Rater guidelines. That means brands with strong E-E-A-T foundations — clear authorship, cited expertise, well-structured schema — are better positioned for Gemini citations than those who’ve been chasing conversational tone for ChatGPT’s browsing model.

For Southeast Asian brands, this matters acutely. Gemini’s rollout across markets like Thailand, Indonesia, and the Philippines is accelerating. Brands that establish entity authority now — through consistent structured content, regional press mentions, and platform-linked profiles — will compound that advantage as Gemini’s local index matures.

Google’s New Agent Architecture Is Not a Footnote

Search Engine Journal’s Roger Montti flagged something that deserves more strategic attention than it’s getting: Google’s introduction of a new AI user agent appears connected to a resource shift from Project Mariner toward a Gemini-native agentic framework. The working theory is that this relates to the broader “OpenClaw” trend — an industry-wide move toward open, interoperable agentic systems.

What this means practically: Google isn’t just answering queries through Gemini. It’s building toward autonomous agents that navigate the web on users’ behalf — reading, comparing, transacting. The sites those agents visit and cite will be the ones with clean crawlability, unambiguous entity signals, and structured data that machines can parse without interpretation.

If your site still relies on implicit brand signals — vibes-based authority, if you will — this shift should feel urgent. Agentic systems don’t infer. They parse.


The Measurement Problem Nobody Wants to Admit

Here’s the part that makes GEO reporting genuinely difficult right now: most teams can’t accurately attribute traffic from AI referrals because the volume is still small enough to get lost in direct or dark social buckets. SE Ranking’s methodology captures a real trend, but as Search Engine Journal contributor Bengu Sarica Dincer argued this week, communicating what data can’t prove is as strategically important as what it can.

For GEO specifically, this means resisting the temptation to over-index on early referral numbers. Gemini’s traffic doubling sounds dramatic; in absolute terms, it may still represent a fraction of organic search volume for most categories. The smarter move is to treat current AI referral data as directional signal, not performance benchmark — and build measurement infrastructure now so you’re not scrambling to attribute it when volume reaches decision-relevant scale.

That means: UTM-tagging all owned content for AI crawler attribution, monitoring brand mention velocity across LLM outputs using tools like Profound or Brandwatch’s AI monitoring layer, and establishing a citation share baseline by category before your competitors do.

What GEO-Ready Content Actually Looks Like in 2026

The practical question isn’t whether to optimise for Gemini — it’s how to do it without abandoning what works for traditional search. The answer, fortunately, is that they’re more aligned than the hype suggests.

Gemini citations favour content that is: specific and attributed (named experts, cited data, datestamped claims), structurally clear (FAQ schema, HowTo markup, clear H2 hierarchies that match likely query patterns), and entity-connected (your brand, authors, and topics linked to Knowledge Graph entities via consistent name-address-phone data, Wikipedia presence, or Wikidata entries).

For multilingual Southeast Asian brands — think a Lazada seller with Indonesian and Thai storefronts, or a regional fintech with separate market landing pages — this means entity consistency across language variants is non-negotiable. Gemini’s multilingual model will conflate or fragment your brand entity if your structured data tells three different stories across three market domains.

One tactical starting point: audit your sameAs schema properties across all market sites. If your Indonesian domain points to a different LinkedIn page than your Thai domain, you have an entity fragmentation problem that no amount of content quality will fully compensate for.


Key Takeaways

  • Gemini’s referral growth reflects its entity-graph advantage — brands with strong E-E-A-T and schema infrastructure are already better positioned for citation.
  • Google’s agentic architecture shift signals that machine-parseable content clarity will matter more, not less, as AI search matures.
  • Build AI referral measurement baselines now; waiting until volume is significant means starting attribution work from scratch under pressure.

The broader provocation worth sitting with: as Gemini consolidates AI referral authority inside Google’s ecosystem, we may be heading toward a world where generative search and traditional search are effectively the same surface — just rendered differently. If that’s true, the brands treating GEO as a separate workstream from SEO are already behind. The question is whether your content infrastructure was built to serve one master or both.


At grzzly, we work with growth teams across Southeast Asia who are navigating exactly this transition — building entity authority, structured content systems, and GEO measurement frameworks that compound across both traditional and generative search. If you’re trying to figure out where Gemini fits in your search strategy for 2026, we’re already deep in that conversation. Let’s talk

Editorial illustration of a grizzly bear quietly rearranging search result cards on a wall while other figures sleep
Illustrated by Mikael Venne
Sneaky Grizzly

Written by

Sneaky Grizzly

Tracking the quiet revolution inside LLM-powered search — where brand mentions, structured semantics, and entity authority rewrite the rules of discoverability before most marketers notice.

Enjoyed this?
Let's talk.

Start a conversation