What is Topic Research?
Topic Research lets you ask AI a question about any of your conversation topics and get an answer in seconds — instead of reading through hundreds of individual chats yourself.
Pick a topic (for example "Shipping Delays" or "Pre-Purchase Questions"), ask something like "What are the most common complaints?" or "Why are customers hesitating to buy?", and the AI reads every conversation in that topic for your selected time period and gives you back a clear summary, the key findings, and suggested follow-up questions.
Topic Research is part of the Deep Research add-on and is available to merchants with an active Deep Research subscription.
Why use it
Understand customers at scale — get the themes inside a topic without opening a single transcript.
Spot problems early — surface recurring complaints, product issues, or points of confusion.
Make confident decisions — base merchandising, support, and product choices on what customers are actually saying.
Save hours — turn what used to be a full afternoon of reading into a few seconds.
How to run a Topic Research query
Go to the Insights menu in your Rep console and open Deep Research.
Set the time period you want to analyze using the timeframe picker on the page. Topic Research uses this date range, so choose it before you open a topic.
Click any topic row (in Top Topics, Your Topics, or AI-Discovered Topics). The Topic Research window opens.
The window shows the topic name, how many conversations it covers for your selected dates, your remaining credit balance, and a few suggested questions to get you started.
Click a suggested question, or type your own question in the input box.
Before you submit, you'll see exactly how many credits the query will use.
Submit your question. The AI reads the conversations and returns your results in a few moments.
What you can ask
Topic Research only answers questions about the customer conversations in the topic you've opened. Ask about patterns, reasons, complaints, requests, or sentiment — for example:
"What are the most common reasons customers reach out about this?"
"Which products do customers ask about most here?"
"What's frustrating customers in these conversations?"
"Are customers satisfied with the answers they're getting?"
Greetings, small talk, or questions that aren't about your conversation data — like "hello," "thank you," or "what is your name" — aren't research questions, so they won't return an answer. If you enter something that isn't a data question, you'll see a short message that it wasn't recognized as a query. Just rephrase it as a question about the topic's conversations and submit again. Because nothing was analyzed, these attempts don't use any credits.
Understanding your results
Each result includes:
A summary — a short, plain-language answer to your question.
Key findings — the main takeaways (up to four), each with a short explanation and the number of conversations that support it.
Follow-up suggestions — one-tap questions to dig deeper, plus you can always type your own.
You can export any result as a PDF at no cost — useful for sharing with your team.
Research history and revisiting answers
Every question you run is saved in the history panel for that topic, so you can come back to past answers anytime. Revisiting a saved result is always free — it doesn't use any credits.
How credits work
Topic Research runs on a simple, transparent model:
1 credit = 1 conversation the AI reads. A query's cost equals the number of conversations in that topic for your selected dates. For example, a topic with 320 conversations costs 320 credits to research.
You get a monthly credit allowance included with your Deep Research subscription. Larger plans include more credits.
Credits are only used when a query succeeds. If a query fails, you are never charged.
Credits refresh every month on your billing cycle date. Unused credits do not roll over.
Credits are shared across everyone on your account.
Revisiting saved (cached) results is free.
You'll always see the cost before you run a query, so there are no surprises.
You may see two credit states:
Running low (amber): your remaining balance is close to the cost of one more query on the current topic. Smaller topics may still be affordable.
Out of credits: you can't start new queries until your monthly refresh — the window shows the refresh date. You can still open and read all of your past results and browse conversations in the meantime.
💡 Tip: a query's cost is the size of the topic for your chosen dates. To research a large topic more cheaply, narrow the date range on the Deep Research page first, then open the topic.
Changing the time period
Topic Research uses the date range set on the Deep Research page. To research a different period, close the window, change the timeframe on the page, and reopen the topic. Each saved result keeps the date range it was originally run with.
Troubleshooting
"Not recognized as a query" / "Not a data question" — your input wasn't a question about your conversation data. Greetings and general questions (like "hello" or "what is your name") won't work. Rephrase it as a question about the conversations in the topic — see "What you can ask" above. These attempts don't use any credits.
"No conversations found" in a topic — there's no conversation data for that topic in the selected dates. Try widening the date range on the Deep Research page.
A query failed — use Try Again. Failed queries are never charged.
I can't start a query / the balance is red — you're out of credits for now. Your balance refreshes on your billing cycle date (shown in the window). Past results stay available.
I don't see Topic Research — it requires an active Deep Research subscription. Check the Deep Research add-on in your Billing page.