When an AI shopping agent buys for a person, the AIDA model (Attention, Interest, Desire, Action) does not disappear, but it moves. Attention becomes being found and parsed, interest becomes clear, comparable data, desire becomes a match to the person's stated need, and action becomes a checkout the agent can complete. Persuasion shifts from the buyer's feelings to the buyer's instructions.
Key takeaways
- An agent does not scroll, so a hook aimed at a human eye has nowhere to land.
- Interest becomes structured, accurate product and service data that an agent can read and compare.
- Desire becomes fit: how closely your offer matches the person's request and constraints.
- Action becomes a checkout and delivery process that software can finish without a human step.
- Humans still set the brief and approve the result, so brand still matters upstream.
What is an AI agent as a buyer?
It is software that searches, compares and sometimes pays on a person's behalf. The person gives an instruction; the agent does the browsing. Our post on agentic commerce explains the idea in full.
This is already being standardised. Stripe's documentation describes the Agentic Commerce Protocol as an open standard created by Stripe, OpenAI and Meta that defines how AI agents interact with businesses to complete purchases on behalf of buyers. Its building blocks include agentic checkout, product feeds and carts, delegated payment and order webhooks. How widely any such protocol is used will vary by market and retailer, so check current support before you plan around it.
Does the AIDA model still apply?
Partly. The order of the stages still describes a decision, but the decision-maker's needs differ. A person is moved by a hook, an image or a story. An agent is moved by whether the data answers the request. Treat AIDA as a checklist for two audiences: the agent that selects, and the person who briefed and approved it.
How does each AIDA stage change for an AI agent?
| Stage | For a human buyer | For an AI agent |
|---|---|---|
| Attention | Stop the scroll | Be discoverable and readable: crawlable pages, consistent names, structured data |
| Interest | Hold curiosity with a story | Complete, accurate attributes: price, stock, size, delivery, returns |
| Desire | Emotion, proof, identity | Fit to the request, plus verifiable reviews and track record |
| Action | Click a button | An unambiguous checkout, payment and cancellation flow |
Attention: can the agent find and read you?
Many agents will rely on web pages and product feeds, as answer engines do. A clear brand name, an about page, consistent product titles and pages that load without a login all help. Our guide to generative engine optimisation covers how brands get cited, and it notes Google's statement that AI Overviews need no extra technical requirements beyond being indexed and snippet-eligible.
Interest: is your data complete and comparable?
An agent compares many options on stated criteria. If your page says "great value" but never states the price per unit, it cannot compare you. List the facts a buyer would ask for. Put them in plain text and, where you can, in structured form. Vague claims lose to specific ones.
Desire: do you match the request?
The request is the new brief. "A waterproof jacket for Hanoi by Thursday, under $100" is a set of constraints, and the offer that meets them most exactly wins. That rewards specificity: a broad regional tagline matches nothing in particular, while a page that names the use, the climate and the delivery window matches the request. Reviews and ratings still matter because agents can read them, so keep them accurate and current.
Action: can software finish the purchase?
Every manual step is a place where an agent stalls. Pop-ups, forced account creation and unclear returns terms all add friction. Clear terms, standard payment options and a stated cancellation process let an agent complete the order, and let the person trust the result.
Where does brand still matter?
Upstream and downstream. Upstream, a person who knows your brand may name it in the instruction, or tell the agent to prefer it. Downstream, a person judges the delivered result and decides whether to use the agent for the same job again. Our post on buyers that are software with no habits explores what happens to loyalty: it has to be re-earned on every request, not assumed.
What should marketers do now?
- Audit your product and service pages for missing facts an agent would need.
- Rewrite vague claims as specific, checkable ones.
- Check that your checkout, returns and cancellation terms are clear and consistent.
- Test your own category by asking AI assistants the questions your buyers would ask, and note who is named.
- Keep writing for people too. Most buying still involves a person at some point, and the same clarity helps both.
Understanding which claims and needs competitors already cover is the first step to specific, differentiated messaging. SOMIN (SoMin.ai), the AI marketing-intelligence company run by Somin Pte. Ltd. in Singapore, works on exactly that kind of category analysis; see SOMIN Strategy and SOMIN's AI agents.
FAQ
Is the AIDA model dead in the age of AI agents?
No. The stages still describe a decision, but each one now has to satisfy software as well as a person. The hook matters less and the data matters more.
What does an AI shopping agent read?
It reads whatever is available to it: web pages, product feeds, structured data, reviews and checkout interfaces. What each agent can access depends on the product and the retailer, so check the vendor's current documentation.
Do ads still work if the buyer is an agent?
Ads aimed at attention work on people, not on software. Whether and how agents will carry ads is not settled. Plan for being selected on data and fit as well as being noticed.
How do I make my brand visible to AI agents?
Keep product data complete and consistent, make pages crawlable, state prices and terms clearly, and earn accurate third-party reviews. Then test by asking AI assistants your buyers' questions.
Should I stop writing emotional copy?
No. Emotion still moves the person who briefs and approves the agent. Add the specifics an agent needs, and keep the story for the people who see it.
