Configure the FAQ Page Agent to generate relevant FAQs from a webpage URL. The agent analyzes the webpage, identifies the primary search queries associated with the page, researches related questions, groups questions with similar intent, and generates the final FAQ using the available webpage content and brand context.
Before you begin
Make sure you have:
A configured Brand Identity.
A webpage URL that you want to use as the source for generating FAQs.
Access to the Content hub.
Navigation
Go to the Content hub.
Expand Agents, select FAQ page agents, and switch to the Library tab.
Hover over the pre-configured FAQ page agent and select Use Agent.
Configure the FAQ Agent
FAQ page agent node
Click the FAQ page agent node to open the Agent details panel.
In Brand identity, select the Brand identity you want to use for the agent.
The Locations are automatically populated based on the selected Brand Identity.
Review the selected locations and click Save.
The Agent details panel also displays the agent's Name, Goals, and Outcomes.
On demand FAQ generation
The On demand FAQ generation trigger starts the workflow when someone requests FAQs for a business or a piece of content.
Collect web pages content for analysis
The Collect web pages content task reads the webpage provided in the Page URL and extracts its content for analysis.
The task uses the Collect webpage content tool, with Website URL mapped to Page URL.
Identify core search queries
The Identify core search queries task reviews the collected webpage content to identify 3–5 search queries that best define the business. These queries are used as the foundation for the subsequent search and FAQ research tasks.
Add context
Use Context to provide additional information that helps the task identify the most relevant search queries for the business.
You can add contexts based on your requirements. For example:
Business Category — provides the business category to help identify search queries relevant to the type of business.
Brand Profile — provides additional business and brand information to help the agent understand what the business offers and how it is positioned.
Add input fields
In the input fields, add the information the task needs from the previous steps to identify the core search queries.
For this task, the configured input fields are:
Scrape content
Page Info
Web page content
These inputs provide the scraped webpage information and page details that the task analyzes to identify the 3–5 search queries that best define the business.
Click Add to include additional input fields as needed for the task.
Configure the user prompt
The User prompt defines the instructions and information provided to the task when it runs. The existing prompt includes the Page Info, Web page content, and Page URL fields used to generate the core search queries.
You can add additional fields to the user prompt based on the requirements of the task. Adding relevant fields can provide the task with more context and help improve Prompt Strength.
The task currently has the following Output field:
Core search queries
This output is used by subsequent tasks in the FAQ workflow.
Prompt Strength is displayed below the prompt and indicates how well the configured prompt is structured based on the information provided.
Click Save after configuring the node.
Search using Perplexity
The Search using Perplexity task checks how an AI answer engine responds to the core search queries identified in the previous step.
This is important because the FAQ content is being created for a webpage that users may discover through AI-powered search. By checking the current responses for the same search queries, the agent can understand what information is already being surfaced by an LLM and use it as a benchmark when generating the FAQ. The goal is to create content that matches or improves on the information already being surfaced.
The task uses the same core search queries identified in the Identify core search queries task as its input.
Find related questions
The Find related questions task finds Google “People also ask” questions related to the search queries identified earlier in the workflow.
People Also Ask results help expand the FAQ beyond the original search queries by identifying related questions that users may ask about the same topic. This gives the agent a broader set of questions to evaluate when selecting the most relevant FAQs for the business.
The task uses the Related questions finder tool to find these questions for a specific search and location.
Rank questions by relevance
The Rank questions by relevance task expands the core search queries into different ways users may realistically search for the same information. It creates variations such as long-tail keywords, voice-search queries, and modifier-based searches, and evaluates them for relevance.
This helps the agent build a broader set of realistic search questions before selecting the final FAQ questions. Each variation is linked back to the source query it was generated from.
Input: Core search queries
Output: Fanout search queries
You can add or modify the fields used in the User prompt to provide the task with additional information relevant to the query expansion.
Select the final question set
The Select the final question set task reviews the questions collected from the different research sources, including customer signals, Perplexity research, Google People Also Ask, core search queries, and the expanded search variations.
It groups questions with the same search intent into a single canonical question and selects the strongest 8–15 question clusters to form the final FAQ set. This prevents duplicate or overlapping questions from being included in the final FAQ and ensures the selected questions are supported by the research collected earlier in the workflow.
The task also ensures that questions are supported by the available input data rather than creating unsupported questions.
Write the final FAQs
The Write the final FAQs task turns the selected question set into clear question-and-answer pairs. It uses the selected questions, webpage facts, brand information, and location context to generate the final FAQ content.
The configured system prompt instructs the agent to generate one answer for each selected question cluster, determine whether each question requires business-specific or general information, lead with a direct answer, keep the answers concise, match the brand voice, and use only supported facts.
Send FAQ to Content hub
This is the final step in the workflow, where the generated FAQ pairs are sent to Content hub to complete the FAQ generation process.
Test the Agent
After configuring the agent, run a test to verify that the workflow runs as expected.
Click the Run test button to start the test.
In Preview settings:
Enter the Website URL you want the agent to analyze.
You can also add any specific topics or requirements you want the generated FAQs to address.
Click Run test to run a dry test and preview the FAQ recommendations the agent generates.
Review the test results
After the test completes, Test details show the outcome of each step in the workflow.
A successful task is marked with a green check.
You can expand the available sections under each task to inspect the corresponding inputs and outputs. For example, expand the FAQ question pool under Select the final question set to review the questions selected for the final FAQ set.
The Write the final FAQs task displays the generated FAQ output.
Send FAQ to Content Hub confirms whether the generated FAQs were successfully sent to Content hub.
Review the test results to verify that the workflow produces the expected questions and FAQ content. Once you are satisfied with the results, click Exit test.
Publish the Agent
After configuring the workflow, you can publish the FAQ Page Agent. The configured agent can then be used to generate FAQs from webpage content through the Content hub.
Next steps
Once you have configured and tested the FAQ Page Agent, you can use it to generate FAQs.
Read: How to How to generate FAQs with the FAQ Page Agent to learn how to generate FAQs using the configured agent.