Chat search
Chat search is the same engine as the guided form, with the difference that you describe what you want instead of filling fields. Underneath, Claude reads your message, extracts a structured ProductSearchRequest, and runs the same search — same SellerScore, same results.
In this guide:
- Open the chat
- Describe what you want
- Watch the inline search trigger
- Continue the conversation
Step 1: Open the chat
From the Product Research home, click Chat (or the chat icon in the FAB on mobile).
Step 2: Describe what you want
Type a message that describes the niche. The AI handles ambiguity well; you don’t have to think in form-field terms.
Examples that produce good searches:
- “women’s bamboo socks under $20 with low competition in the UK”
- “reusable coffee filters in the US, FBA only, around $15–$30, prefer high-trend products”
- “yoga mat straps in DE that aren’t already saturated, want at least 30% margin”
- “new vape accessory niches with rising trend” (the AI may push back if it can’t satisfy a constraint)
The AI translates that into a structured search and runs it inline.
Step 3: Watch the inline search trigger
When the AI decides your message is a search request (intent extraction), you’ll see a sequence:
- Message acknowledged. A short ack from the AI (“Searching for women’s bamboo socks under $20 in the UK…”).
- Search triggered. A
search_triggeredevent with the parameters the AI extracted. - Search completed. The top 5 results render inline in the chat, with sub-score chips.
- Streamed analysis. The AI continues with a brief analysis of the results.
The triggered search costs 1 credit — the same as a guided-form search. The chat itself is free.
Step 4: Continue the conversation
You can refine without starting over:
- “Tighten that to 25%+ margin.”
- “Also show me CA results.”
- “What’s the catch with the top one?”
Each refinement either runs a new search (charged 1 credit) or just reasons over the existing results (free). The AI tells you which it’s doing.
When chat doesn’t trigger a search
If your message is conversational rather than a search request — “explain how the SellerScore works”, “what’s a good margin for FBA?” — Claude responds without triggering a search. No credits charged.
How does Claude know? Two tools: extract_search_params (for search messages) and no_search_intent (for conversational ones). The AI picks one based on what you wrote.
Tips for getting useful searches out of chat
- Be specific about marketplace (“in the UK”, “for the US store”). The AI defaults to your primary marketplace if you don’t say.
- Give a price band (“under $30”, “between $15 and $50”). The AI doesn’t guess.
- Mention what you’re avoiding (“not too saturated”, “prefer trending products”). Becomes part of the score reasoning.
- Ask one search question per message. Multi-question messages confuse intent extraction.
When to prefer guided over chat
Chat shines for exploration and nuance. Guided shines for repeatable, parameterized research where you want exact control. If you find yourself fighting the AI to nail the parameters, switch to guided — same engine, fewer surprises.