Original post by Wil Reynolds and Andrea Haley
30-second rundown
Key learnings:
- AI can rebuild the shortlist: AI answers can change which brands a shopper considers before that shopper visits a website.
- The buyer speaks in constraints: Create content that answers concrete buying situations, such as household size, installation limits, budget, or use case.
- Test the task and the answer: Test the same buyer tasks repeatedly and track which claims and sources cause brands to enter or leave the shortlist.
This week: Pick one product category and write five buyer questions based on real constraints. Ask five colleagues to name brands before using AI, run the questions, record their final shortlist, and note which evidence changed their minds.
Coming 3 months: Build a monthly task-based test using the same prompts and a consistent participant group. Turn repeated buyer concerns into demonstrations, comparisons, and use-case pages, then monitor whether your brand appears more often for the situations where it genuinely fits. Keep the sample limitations visible.
Wil Reynolds and Andrea Haley put ordinary shoppers in front of AI and watched familiar brand names get stripped for parts.
The machine did not merely answer a question. It changed which companies people considered, sometimes in the space of one conversation, with all the ceremony of a roadside robbery.
1. AI Can Rebuild the Shortlist
Seer ran 84 shopping sessions with 28 participants completing three tasks each. In the water-heater exercise, General Electric fell from 18 initial mentions to 8 final choices, Whirlpool fell from 9 to 2, Rheem rose from 3 to 9, and Delta rose from zero to 5. The sample is small and directional, not a market forecast, but it shows how an answer can rearrange consideration before a customer ever reaches a product page.
2. The Buyer Speaks in Constraints
People asked about family size, bathrooms, shower use, installation limits, and other practical conditions. Brands that explain those real situations give AI systems material to use when comparing options. A page stuffed with category phrases but silent about the buyer’s problem is a chrome-plated engine with no fuel.
3. Research the Task, Then Test the Answer
Seer recommends starting with the job a customer is trying to complete and using sources such as Google autocomplete to collect the language around it. Turn those needs into repeatable prompts, record which brands appear before and after the AI conversation, and inspect the sources and claims that drove the change. Monitor communities such as Reddit for recurring concerns, but use them as research evidence rather than a dumping ground for synthetic promotion.
The warning is not that every prompt destroys years of brand building. It is that brand memory now enters a live interrogation room where practical evidence can overturn familiarity.
Show the machine how your product survives the buyer’s actual life, or a lesser-known rival may walk out wearing your customers.
