This guide tells how Genezio runs conversations with answer engines and how it changes the responses into results. Conversations are the core mechanism of Genezio: with them, Genezio measures what answer engines say about your brand and your competitors.

​
How Genezio runs conversations each day

Genezio runs conversations automatically on a daily schedule. Each day, Genezio runs each scenario with each selected answer engine. For example, you have:
  • 10 scenarios
  • 4 selected answer engines
Then, Genezio runs 40 conversations each day. Genezio stores and analyzes each of these conversations separately.

​
How the persona changes the conversation

The persona of the topic has an effect on most conversations. When a topic has a persona, Genezio runs the conversations of that topic like this:
  • Genezio writes the interaction with the answer engine as that persona.
  • Genezio runs the conversation in the language of the persona.
  • Genezio runs the interaction from the geographic location of the persona.
Thus, the answer engine responds as if a real user from that context asked the question. Different personas can get different answers from the same answer engine. For example, a student, a startup founder, and a corporate buyer can ask about the same product category, and get different recommendations. Introspector topics are an exception. They can run with a persona or without a persona. When the topic has no persona, the conversation only asks the answer engine to describe the brand directly, with no audience overlay. To learn when to use each option, refer to Introspector Agent.

​
How Genezio makes multi-turn conversations

A two-agent conversation driver controls the multi-turn conversations. It has a planner agent and a replier agent that operate as a pair:
  • The planner selects the next question of the simulated user.
  • The replier writes the messages of the user in the exchange with the answer engine.
Two agents make more realistic and natural multi-turn conversations than one driver. Each follow-up question is similar to a question of a real person, and not to a scripted continuation. The two-agent driver applies to the multi-step agent types: Recommender, Comparer, and Introspector. These agent types make an answer over many turns. Prompter is the exception. It sends its one prompt without changes, and the planner and the replier do not operate.

​
Why conversations run each day

Answer engines change continuously. Their answers can change because:
  • The providers update the models
  • The sources on the web change
  • New content appears
  • The retrieval system changes
When Genezio runs conversations each day, it can monitor how the answers change over time. It can also find changes in AI Recommendations and Visibility.

​
What occurs during a conversation

When a conversation runs, Genezio does these steps:
  1. Genezio selects a scenario.
  2. Genezio runs the conversation with a selected answer engine.
  3. Genezio records the full response of the answer engine.
  4. Genezio analyzes the response and extracts structured data.
This analysis gives the key signals that the platform uses everywhere.

​
Data from conversations

Each conversation gives many layers of analysis.

​
Query fanouts

Query fanouts are the additional searches that the answer engine does internally when it makes an answer. These follow-up searches show how the answer engine examines the topic. For more information, refer to Query fanouts explained.

​
Citations

Citations are the web pages (sources) that the answer engine used as references for the answer. These sources help to explain where the information in the response came from.

​
Perceptions

Perceptions are the separate claims that Genezio extracts from the response of the answer engine. With perceptions, Genezio evaluates the accuracy of what answer engines say about your brand. Genezio also compares the narratives across conversations.

​
Competitors

Genezio automatically finds the brands that the same response mentions together with your brand. Genezio monitors these brands as competitors.

​
Competitive insights (Comparer conversations)

When you run Comparer conversations, Genezio analyzes how answer engines compare your brand with specific competitors. With these conversations, you can:
  • Understand the strengths and the weaknesses that answer engines connect to each brand
  • Find opportunities where competitors are weak
  • Find threats where competitors have a better position
  • Make SWOT-style insights for your competitive strategy
Genezio extracts the SWOT analysis automatically and shows it in these places:
  • In the text of a Comparer conversation, when you open it in the conversation drawer
  • Aggregated at the scenario level and at the topic level
  • Aggregated for your full landscape in the Competitors -> SWOT view
Thus, you get a clear picture of how answer engines position your brand in relation to the alternatives. You also see where to put your content effort. For more information, refer to SWOT Analysis.

​
Reports in the language of your brand

Genezio writes the extracted perceptions, the SWOT, and the AI insights in the language of the brand, not only in English. Teams in markets that do not use English get readable reports that they can share without translation. You set the language of a brand during the brand setup.

​
Examine conversations

You can examine each conversation in the Conversation Detail View. This view shows:
  • The full interaction with the answer engine
  • The prompts that Genezio sent to the model
  • The responses
  • The query fanouts that Genezio found
  • The extracted citations
  • The extracted perceptions
With this view, teams can understand exactly why a brand appeared or did not appear in an answer.

​
Product-aware transcripts

The transcripts are product-aware. Genezio tags and highlights each product that an answer mentions. Genezio also shows the tables and the maps of the response as tables and maps, not as raw text. Thus, you can scroll a conversation and find the exact moment when the answer engine recommended a product. You can also find the moment when it named a product and then selected a different one. For brands with Shopping enabled, the transcripts are the evidence layer below the product metrics.

​
Run conversations again

Usually, Genezio runs conversations automatically each day. But you can also run them again when:
  • You update a scenario
  • You add a scenario
  • You change a topic
Thus, you can quickly test new scenarios and see how answer engines respond. To test a conversation without an effect on your metrics, use the Playground.

​
Next steps

To learn how Genezio changes the responses of answer engines into structured data, read these pages:

​
In the API

The public API reads conversations. Refer to List conversations.