EU Classifies ChatGPT as Search Engine, Forcing AI Citation into Marketing Budgets
Executive Summary
The European Commission's VLOSE designation of ChatGPT under the Digital Services Act converts AI search from a pilot experiment into a procurement-ready channel — overnight. New citation data from Azoma reveals ChatGPT and Google Gemini invert their source preferences almost exactly, meaning a single GEO strategy optimises for one agent and actively misaligns every other. Commission a per-agent citation audit across your top five categories before October 2026 or cede the baseline to competitors who move first.
The Signal
The European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act on August 31, placing it in the same regulatory category as Google and Bing. The classification applies to a platform now reporting over one billion weekly active users globally. Citation behaviour across AI shopping agents varies sharply by platform: ChatGPT draws 41% of citations from earned media and 37% from retailers; Google Gemini inverts that at 41% retailer and 37% earned media; Alexa for Shopping skews heavily toward affiliate sources at 73%. Azoma, which tracks citation patterns across ChatGPT, Google Gemini, Amazon Rufus, Alexa for Shopping, and Walmart Sparky, published the citation breakdown from Q2 2026 analysis of millions of shopping agent responses alongside the regulatory news.
What Changed
Brands can now track AI citation share-of-voice at prompt level across seven distinct AI shopping agents simultaneously, with per-agent source breakdowns showing exactly which content types — earned media, retailer pages, brand.com, affiliate — each agent favours in a given category. This makes AI visibility measurable and addressable as a discrete channel, distinct from traditional search rank, and optimisable through targeted content and product-data workflows rather than generic SEO.
Why It Matters
The EU's classification of ChatGPT as a Very Large Online Search Engine is less a regulatory milestone than a forcing function — it compels brands to treat AI agent visibility as a measurable, budgetable channel rather than an experimental side project. That shift in institutional framing unlocks procurement decisions that have been stalled in "pilot" status for the better part of a year. The deeper commercial implication is what the citation data exposes: there is no unified AI search audience. ChatGPT and Google Gemini serve near-identical user demographics but invert their source preferences almost exactly — 41% earned media versus 41% retailer, flipped. Alexa for Shopping operates in a different universe entirely, where affiliate content captures nearly three-quarters of citations. A brand that has built its GEO strategy around one agent has, at best, solved for one-third of the AI discovery surface. At worst, it has optimised away from the sources that actually drive citations on the platforms where its category converts. What becomes obsolete is the single-channel GEO brief — the content team tasked with "getting us into AI answers" without specifying which agent, which prompt cluster, or which source type. That approach produces work that is visible nowhere in particular. What becomes newly possible is per-agent content arbitrage: identifying that your category is underserved in earned media on ChatGPT while competitors have saturated retailer pages on Gemini, then closing the gap before the platform's citation patterns harden into convention. The underlying pressure is competitive clock speed. Citation hierarchies across these agents are still forming. The brands that map them now set the baseline everyone else will be chasing in Q1 2027.
Marketing Impact
media
Media planners must now budget AI agent visibility as a discrete line item alongside paid search. The per-agent citation breakdown makes this defensible: ChatGPT and Gemini invert source preferences almost exactly, so channel allocation decisions that ignore agent-level data are structurally misallocating spend against an audience that is actively shifting discovery behaviour.
ecommerce
Product discovery on Alexa for Shopping runs 73% through affiliate sources — a citation pattern that bypasses brand.com and retailer PDPs almost entirely. Ecommerce teams that have not mapped their affiliate content footprint against agent citation behaviour are invisible on the fastest-growing commerce surface in the Amazon ecosystem.
martech
The EU designation turns AI citation tracking from a nice-to-have experiment into a procurement-ready requirement. Martech stacks that cannot report prompt-level share-of-voice across ChatGPT, Gemini, Rufus, Alexa for Shopping, and Walmart Sparky will face replacement pressure in Q4 2026 budget cycles, not gradual phase-out.
product marketing
Per-agent source analysis exposes which content type — earned media, retailer copy, brand.com, affiliate — earns citations in a given category on a given agent. Product marketing teams can now run content arbitrage: identifying gaps where competitors dominate one source type on one agent and closing them before citation patterns harden into convention.
The Exploit
Opportunity
Brands can now audit their AI citation share-of-voice across ChatGPT, Google Gemini, Amazon Rufus, Alexa for Shopping, and Walmart Sparky simultaneously using Azoma's per-agent source breakdowns. The arbitrage window is real: category citation hierarchies are still forming, and a focused 90-day content sprint targeting underserved source types — earned media on ChatGPT, retailer data on Gemini, affiliate alignment on Alexa — can capture positions competitors have not yet identified exist.
Risk
Agent citation patterns are still fluid — content optimised for current source preferences could lose relevance if platform behaviour shifts post-DSA compliance changes, creating stranded content investment across formats that no longer convert.
The Move
Commission a per-agent citation audit across your top five product categories before October 2026; use Azoma or equivalent to identify where your brand is absent by source type, then brief content and retail media teams against the specific gaps — success checkpoint is measurable citation share gain on at least two agents by end of Q1 2027. Own it: VP of Performance Marketing or Head of GEO, not SEO.
First-Mover Advantage
Gains
Early movers establish citation presence before platform algorithms calcify around incumbent sources. Brands that close specific per-agent gaps in Q4 2026 set the share-of-voice baseline every competitor benchmarks against in Q1 2027 planning.
Risks
Agent citation patterns are still fluid — content optimised for current source preferences could lose relevance if platform behaviour shifts post-DSA compliance changes, creating stranded content investment across formats that no longer convert.
Window
The window stays open through Q1 2027, closing when major GEO platforms commoditise per-agent measurement and agency holding companies standardise the playbook — expect that inflection around mid-2027.
Winners & Losers
Winners↑
Multi-agent AI visibility and commerce optimisation platforms
The EU's VLOSE designation converts AI search from an experimental budget line into a procurement-ready channel, and the only vendors positioned to capture that spend are those with prompt-level tracking across all major agents simultaneously. Azoma's seven-agent coverage is the current benchmark, but the designation will accelerate competitive entry — platforms with cross-agent citation analytics should move fast to lock enterprise contracts before the category crowds. The window to establish category leadership is Q4 2026.
Earned media and PR content teams with product category expertise
ChatGPT's citation mix — 41% earned media — means editorial coverage, review placements, and authoritative third-party content function as direct GEO inputs on the world's largest AI platform. Teams that can generate category-credible earned coverage at scale now have a direct line into ChatGPT recommendations that retailer-page optimisation cannot replicate. The play is to reframe earned media output as measurable AI citation fuel and pitch it to CMOs with share-of-voice data to back the ask.
Retail media networks operating AI shopping agents with proprietary citation data
Amazon and Walmart control both the AI shopping agent and the underlying product data that agent cites — a structural advantage that independent brands cannot replicate. With Alexa for Shopping drawing 73% of citations from affiliate sources, retail media partners who understand that citation architecture can sell placement in a way that is invisible to brands optimising through generic GEO tools. These networks should productise their citation data as a premium intelligence layer on top of existing retail media inventory.
Brand content and product data operations teams in regulated or high-consideration categories
Categories where trust signals drive purchase — health, finance, home improvement — are precisely where AI agents will weight authoritative source types most heavily, and the citation data now exists to prove which source types win per agent. Brands with the operational infrastructure to maintain clean, structured product data across retailer feeds and authoritative earned coverage across publisher networks are positioned to dominate AI citations in these categories before competitors understand the mechanism. The immediate action is an audit of product data quality and earned media coverage mapped against the Azoma citation breakdowns for their category.
Losers↓
Traditional SEO agencies without multi-agent GEO capability
The VLOSE designation accelerates the institutional legitimacy of AI search as a channel, which directly exposes the gap between blue-link rank optimisation and AI citation optimisation — a gap these agencies cannot close with existing tooling or methodology. Client conversations that have been deferred as 'AI is still experimental' will now land on procurement desks with regulatory framing attached, and agencies without per-agent citation tracking have nothing credible to sell into that conversation. The defensive move is rapid partnership or acquisition of GEO tooling before Q1 2027 planning cycles close.
Brands running a single undifferentiated GEO strategy across all AI platforms
The citation data makes clear that optimising for one agent's source preferences actively misaligns a brand's content posture for every other agent — a Gemini-first retailer-page strategy, for instance, produces content that is structurally deprioritised by ChatGPT's earned-media weighting. Brands with a single GEO brief and no per-agent measurement infrastructure are not just underperforming; they are potentially compounding the problem with each content investment. The corrective is immediate category-level citation benchmarking across agents before committing further budget to the existing strategy.
Affiliate publishers and comparison sites dependent on Alexa for Shopping traffic
Alexa for Shopping's 73% affiliate citation rate looks like a windfall for affiliate publishers now, but regulatory scrutiny triggered by the DSA classification will pressure Amazon to demonstrate source diversity and reduce the appearance of commercially-motivated citation bias. Any compliance response that normalises Alexa's citation mix toward the industry average would eliminate a disproportionate share of affiliate-driven discovery traffic with little warning. Affiliate publishers should treat current Alexa referral volumes as structurally fragile and accelerate diversification into earned media placements that perform across multiple agents.
Strategic Outlook
The EU designation does for GEO what GDPR did for data governance: it converts organisational reluctance into institutional obligation. Procurement cycles that have been stalled at proof-of-concept will move to platform selection in Q4 2026, accelerating consolidation among the handful of tools — Azoma and its direct competitors — that can demonstrate genuine multi-agent, prompt-level tracking rather than dashboard-level approximations. Expect the major analytics platforms and commerce clouds to respond with acqui-hire or feature integration plays by Q1 2027, compressing the window for purpose-built GEO vendors to establish category leadership. The citation data itself becomes a durable competitive asset: brands that build proprietary prompt libraries and citation benchmarks now will own category baselines that late entrants will have to pay to replicate. The structural risk is fragmentation — five agents with meaningfully different citation logic means five optimisation workstreams, and most marketing organisations are not staffed for that without tooling that collapses the complexity.