In March, across ChatGPT and Gemini:
41-46% of users were researching products
31-37% of users compared product prices
25-34% of users actively shopped
And this was when both platforms sat at roughly 400M users (ChatGPT weekly, Gemini monthly).
They're now at 800M weekly (GPT) & 650M monthly (Gemini).
Apply those shopping behaviour ranges to today's user base. You're looking at billions of people starting their purchase journey inside an LLM. And those March percentages have only moved in one direction.
This is where the transaction is going. Here's what you actually need to do about it:
1. Treat AI Agents as Your Number One Customer
No more blocking bot traffic indiscriminately. Every legitimate shopping agent excluded is revenue being handed to competitors.
Your product content strategy needs to flip:
Humans need 3 bullet points. LLMs need thousands of data points per product
Full technical specs, not marketing copy
Structured data feeds optimised for machine readability
Content in consumer lexicon, not brand speak
Every product attribute exposed, even if it feels risky
Your website becomes an AI-readable space and the source of truth for agents:
Easy to crawl, cache, and extract structured data
Focus on GEO (Gen AI Engine Optimisation), surface favourably in answer engine results
2. Get Agnostic on Checkout Infrastructure
The payment protocol landscape is fragmenting: ACP, AP2, A2A, Visa Intelligent Commerce, Mastercard Agent Pay. You cannot build point solutions for each one.
What you need:
Infrastructure platforms that act as a shock absorber between protocols
One integration supporting multiple commerce protocols
Abstracted PCI compliance complexity
Future-proof now and build an open API endpoint layer; not custom integrations you rebuild every 18 months
The shift from authentication to authorisation when agents act on behalf of humans:
Payment systems must verify intent (authorisation) + identity (authentication)
Capture intent signals in non-deterministic environments
Turn text prompts into deterministic purchases
Prepare for marketplace models:
OpenAI and Google are enabling hosted checkouts
Feed inventory to multiple answer engines
Handle orders coming back from platforms you don't control
Build for one-to-many relationships
3. Build an 'AI-native' Company (Internal AI First)
Before consumer-facing agents, fix your back office. Internal AI delivers immediate margin improvement:
Generative AI for coding and quality control
Predictive demand forecasting with hyper-relevance
Supply chain optimisation that moves from insights to predictions
Inventory management that anticipates, not reacts
This is the operational agility that lets you compete when transactions fragment across platforms. Win on margin before you compete on experience.
4. Make Loyalty and Personalisation Your Defensive Moat
In agent-mediated commerce, loyalty programmes become negotiation criteria:
Agents apply points, discounts, special offers on behalf of consumers
No loyalty programme = commodity price comparison
Loyalty value becomes the tie-breaker at similar prices
Enable advanced personalisation that agents can leverage:
Preferred credit card rewards (miles vs cashback)
Buy now, pay later preferences
Individual loyalty status and redemption options
Real-time negotiation capabilities
Permissioned data access in exchange for discounts
The Operational Reality
This isn't a 2027 problem. The user base is already there. The shopping behaviour is already happening. The only question is whether merchants and businesses will build for this reality or will only optimise for yesterday's traffic patterns.
The retailers who win this shift are treating it like infrastructure replacement, not marketing innovation. Because that's exactly what it is.
Get in touch if you want to learn more about how we're preparing customers for the agentic commerce shift.
Written by Mike ✌
Passionate about all things AI, emerging tech and start-ups, Mike is the Founder of The AI Corner.
Subscribe to The AI Corner
