Manifest allows merchants to understand where AI responses are coming from and offers tools to correct or improve those responses. This not only enhances the accuracy of AI interactions but also allows you to maintain control over your AI’s behaviour and knowledge base.
Message Source
The Message Source helps you trace the origin of an AI-generated response. Whether the answer is pulled from FAQs, data sources, product content, or a specific feature setting (like Add to Cart, Quiz, Agent Handover, etc.), this section gives you full transparency—and the ability to edit.
You can view and update the source data directly via inline redirection options provided.
When a Response Is Backed by Data Sources:
You’ll see the following sections when the intent has data or a satisfactory answer:
Intent
Shows the identified intent for the user’s query.Description
Explains how to update the data:
"If any of the information is incorrect, update the original data source and sync it again. Alternatively, you can make local corrections directly in Manifest."Content Source
Displays associated data sources such as FAQs, fetched links, manual links, and text responses.
Editable for everything except PDFs and training templates.
Redirect buttons let you quickly open the Assistant Readiness page for training edits or the Data Sources page for PDF updates.
If a data source has been deleted, a tooltip will appear: "This source is deleted."
Product Source
Shown when the query relates to product details (e.g., search, specs, similar items).
Lists relevant products, their descriptions, FAQs, and metafields.
Includes redirection to the Product Table for quick edits.
Training Insights
For product-related queries, it explains:
"Products are recommended by matching queries with product names, descriptions, tags, and collections. Add relevant keywords and tags to your products to improve results."For other queries, it shows the number of FAQs available for that intent.
Action buttons like "Add relevant data" or "Add FAQ" will be visible.
When the Response Has No Data Sources and Is Unsatisfactory:
You’ll only see:
Intent
A note: "No data available to answer this query. Add an FAQ to address it in the future."
In this case, merchants can add an FAQ directly using the Add FAQ button in the insufficient data modal.
How to View the Message Source:
Go to the Inbox in the Manifest dashboard.
Select any conversation.
Click "Message Source" on any AI response.
View where the AI fetched its answer from and make necessary edits.
How to Edit a Message Source:
Go to Inbox → Open a conversation → Click Message Source.
Under Content Source, find the listed data (text, fetched links, FAQs, etc.).
Click the Edit icon beside any source.
A side modal or FAQ modal will appear based on content type.
Make changes and click Save or Save & Exit.
Edits will reflect in both the inbox and the respective data source (e.g., FAQs).
Correct Response
Correct Response lets merchants correct or rewrite AI answers directly from the inbox. Once corrected, the response is saved as an FAQ so that similar queries in the future receive the improved answer.
Key Highlights:
Corrected answers appear as a private note within the conversation.
You can edit a previously corrected response, and the original FAQ will be updated.
Not all responses can be corrected — e.g., product search result summaries, agent handover prompts, order tracking, feedback collection, email transcript messages, etc.
Correct response is not shown for replies with no data source or unsatisfactory answers.
You can also test and correct responses in the Playground while prototyping conversations.
How to Use Correct Response:
Navigate to Inbox in Manifest.
Select a conversation and click Correct Response on an AI-generated message.
An FAQ modal will open with pre-filled values:
Question = User’s query
Answer = AI’s original reply
Intent and related Product (if applicable)
Edit the content as needed and click Save.
The corrected response will now be stored as a new FAQ entry.
FAQs
1) What is the Message Source in Manifest?
It shows where the AI picked up the information for a specific response — including data sources, product content, or feature settings — and allows merchants to edit those sources.
2) Can I edit the response directly from the message source?
Yes, you can click the Edit icon in the message source section to modify FAQs, fetched links, manual entries, and text responses. Changes will reflect immediately in both the inbox and data source.
3) Why don’t I see the Correct Response option on some messages?
Correct Response is not available for messages related to product search summaries, agent handovers, feedback, lead capture, email transcript notifications, or when the AI has no data source for the answer.
4) Can I edit a response that was already corrected?
Yes. You can edit the previously corrected response, and the same FAQ entry will be updated accordingly.
5) What is the Correct Response feature used for?
It allows you to correct an AI response and save it as a new FAQ. This ensures future responses for the same or similar queries are more accurate.
For further assistance or to raise feature requests related to Message Source & Correct Response please contact [email protected].