Lesson 1: How People Research and Buy Through AI Assistants

Why this matters now

For twenty years the buying journey ran through a search box, a list of links and your website. Now a growing share of that journey happens inside a conversation. A customer types a question into ChatGPT, Microsoft Copilot, Gemini, Perplexity or Google AI Mode, gets a summarised answer, asks follow-ups, and sometimes buys without visiting a website at all.

This lesson maps that journey and shows where a business can appear in it. The status notes below were read on 20 October 2026. This field changes every month, so treat every status as a snapshot.

The five stages of an assistant-led journey

Think of the assistant as a knowledgeable shop assistant who goes with the customer from the first question to the checkout. At each stage the customer asks a different kind of question:

1. Discover. The customer describes a need, not a product. Example: "I need a waterproof jacket for walking in Scotland in autumn."
2. Research and shortlist. The assistant suggests options. The customer narrows them down: "Which of these are under £150?"
3. Compare. The customer asks for side-by-side detail: "Compare the first two on weight, warranty and reviews."
4. Decide and buy. The customer acts: "Where can I buy the second one?" or completes the purchase inside the chat, where that is available.
5. After purchase. The customer comes back for help: "How do I reproof this jacket?" This stage shapes repeat purchases and reviews.

One difference from search matters most: the customer rarely sees ten blue links. They see one written answer. If your business is not part of that answer, or of a surface next to it, you are often not in the journey at all.

The four surfaces where a business can appear

At each stage, a business can show up in up to four kinds of surface:

• Answers. The assistant mentions or cites your brand, product or page in its reply. You do not pay for this, and you cannot fully control it.
• Ads. These are paid placements shown alongside the answer. OpenAI's help centre says ads in ChatGPT may appear for users on the Free and Go plans, below the response, labelled as sponsored. They are not shown on Plus, Pro, Business, Enterprise or Edu accounts, or to users identified as under 18.
• Apps and connectors. These are branded tools that run inside the assistant. Apps in ChatGPT are built with OpenAI's Apps SDK, which is based on the Model Context Protocol (MCP). Since December 2025 developers have been able to submit apps to an in-chat directory.
• Checkout. The purchase is completed in the conversation. Where this is available today:
  • Microsoft Copilot Checkout: rolling out in the US with partners including PayPal, Shopify and Stripe.
  • Google AI Mode checkout: runs on Google's Universal Commerce Protocol (UCP), and the retailer stays the seller of record.
  • Perplexity: offers PayPal-powered purchasing for US users.
  • ChatGPT: OpenAI ended in-chat Instant Checkout in March 2026 and moved towards retailer apps and handing the buyer off to the merchant's own site.

How the surfaces line up with the stages:

Task 1: Sketch your customer's assistant journey

DO this now, in a document or on a whiteboard:

1. Write the five stages across the top of a page.
2. Under each stage, write three questions a real customer might type into an assistant. Use their words, not your product names.
3. Type each question into at least two assistants, for example ChatGPT and Google AI Mode.
4. For each answer, note whether your brand appears, which competitors appear, and which surface they appear in.
5. Mark every gap where a competitor appears and you do not.

Task 2: Find out which journeys matter, using your own data

A journey map is a guess until your data confirms it. DO the following:

1. Referral data. In your analytics tool, filter traffic by referring domain. Look for chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. Note the volume, the trend and the pages people land on.
2. Search data. In Google Search Console, review long, question-style queries. Be aware of a limit: Google includes traffic from AI Overviews and AI Mode in the overall "Web" search type in the Performance report. It does not report that traffic separately.
3. Sales data. Add an "AI assistant" option to any "How did you hear about us?" field at checkout, on forms and in sales calls.
4. Compare. Rank the assistants by referral volume and conversion. The top one or two are where your effort should start.

Expect the numbers to be small and incomplete. When a customer buys inside a chat, or remembers an answer and visits you later, there is often no click to track. A later level covers measurement in depth.

The four channel types in this course

Every later level builds on one of four channel types:

1. Assistant ads: paid placements in ChatGPT, Copilot, Perplexity and Google's AI experiences.
2. Apps and connectors: branded tools that run inside assistants.
3. Agentic commerce: product feeds, in-chat checkout and agent payment protocols.
4. Conversational content: content written so assistants understand it, cite it and answer with it correctly.

Readiness checklist

• Five-stage journey sketched, with three real customer questions per stage
• Questions tested in at least two assistants, with gaps marked
• AI assistant referral domains filtered in analytics
• "AI assistant" option added to attribution questions
• Top one or two assistants for your customers identified

Key takeaways

• An assistant-led journey has five stages: discover, research and shortlist, compare, decide and buy, and after purchase.
• A business can appear through four surfaces: answers, ads, apps and checkout.
• In-chat checkout varies by assistant and changes quickly. ChatGPT ended Instant Checkout in March 2026, while Copilot, Google AI Mode and Perplexity offer checkout in the US.
• Your referral, search and sales data, not industry headlines, should decide which assistants you prioritise.
• This course covers four channel types: assistant ads, apps and connectors, agentic commerce and conversational content.