Understanding Clay's Data Levels: Research3 vs. Main Body Text

Richard X. · 2025-06-22T04:19:08.375Z

Look at the image attached. When I select a column for Clay to reference using "/", I can insert the "Research3" (I'm guessing this is all content produced by that enrichment) or I can just insert the "main_body_text" which is really all that I need. 2 questions: (1) What is the difference between the two? (2) If I were asking AI to summarize or generate an email based off content in this enrichment, which level would produce less-hallucinatory, more accurate responses? The "Research3" or "main_body_text". (3) Whenever I regenerate a an AI prompt, if I had used "main-body_text", Clay automatically reverts the enrichment level back to "Research3" - why?

Understanding Clay's Data Levels: Research3 vs. Main Body Text | Clay

8 comments

· Sorted by Oldest

Channeled
APP
·
·
** Difference: "Research3" likely refers to the entire output of a research enrichment, which may include multiple fields or structured data, while "main_body_text" is typically just the core content or main text extracted from a source, such as the body of a web page or article. "main_body_text" is the raw material, whereas "Research3" may include additional context or metadata alongside the main content. ** Accuracy for AI Summarization/Email Generation: Using "main_body_text" generally yields more accurate and less hallucinatory AI responses for summarizing or generating emails, since it provides focused, relevant content without extra noise. Feeding the AI only the necessary text helps maintain control and reduces the chance of errors or irrelevant information being included. ** *Why Clay Reverts to "Research3": Clay may default back to the enrichment-level field ("Research3") when regenerating prompts to ensure consistency or because the system is designed to reference the top-level enrichment by default. This behavior helps maintain a standard reference point, but if you need only "main_body_text, " you may need to reselect it after each regeneration.
Sources: Deep Dive: Clay Scrapers [Limitless Research] - Clay University Transform: Clean, Structure, and Segment Data with AI - Clay University Intro to AI Snippets (Copywriting) [Clay 101] - Clay University
Richard X.
·
·
^ oof... Clay S. do we think we can stop the auto-defaulting if this is case? It's a lot of manual work telling the AI prompt to reference just the "main_body_content" each time after a regeneration. Lmk if I'm misunderstanding or if there's a better way to reference!
Channeled
APP
·
·
The documentation doesn't cover specific settings to prevent Clay from auto-defaulting to enrichment-level fields when regenerating AI prompts. The available guidance focuses on verifying that dynamic prompt inputs are correctly linked to appropriate columns after using the AI help tool. When configuring AI enrichments, you can verify that dynamic prompt inputs are correctly linked to the appropriate columns - this step is crucial when you have multiple columns connected to the same endpoint. Are you experiencing this auto-defaulting behavior specifically when using the "Generate with prompt" feature in the AI help tool, or does it happen with other types of AI prompt regeneration as well?
Sources: Referencing dynamic data from other columns | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University
Richard X.
·
·
yes, "Generate with prompt"
Channeled
APP
·
·
The documentation doesn't include specific settings to prevent Clay from auto-defaulting to enrichment-level fields when using "Generate with prompt." The available guidance focuses on verifying that dynamic prompt inputs are correctly linked to appropriate columns after using the AI help tool. When configuring AI enrichments with "Generate with prompt, " you need to verify that dynamic prompt inputs are correctly linked to the appropriate columns - this step is crucial when you have multiple columns connected to the same endpoint. Are you finding that the auto-defaulting happens immediately after clicking "Generate with prompt, " or does it occur when you make other changes to the prompt configuration?
Sources: Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University
Piotr M.
·
·
Richard X. What can really help is a function that generates prompts with AI. It uses the SPICED framework and also produces results automatically, without any additional parameters if you don’t want them.
LuisArturo
·
·
Hey there Richard sorry for the delay here, taking a look at this 1. The difference here is that by entering in the "Research3" all of the values inside of the cell will be used here vs just the single value that was selected in "main_body_text" 2. If you wish to have answers that would cause the AI to not hallucinate as much then entering in the "main_body_text" by itself would be best here. 3. Could you share the table link here so that we can check.
Channeled
APP
·
·

Channeled
APP
·
·
** Difference: "Research3" likely refers to the entire output of a research enrichment, which may include multiple fields or structured data, while "main_body_text" is typically just the core content or main text extracted from a source, such as the body of a web page or article. "main_body_text" is the raw material, whereas "Research3" may include additional context or metadata alongside the main content. ** Accuracy for AI Summarization/Email Generation: Using "main_body_text" generally yields more accurate and less hallucinatory AI responses for summarizing or generating emails, since it provides focused, relevant content without extra noise. Feeding the AI only the necessary text helps maintain control and reduces the chance of errors or irrelevant information being included. ** *Why Clay Reverts to "Research3": Clay may default back to the enrichment-level field ("Research3") when regenerating prompts to ensure consistency or because the system is designed to reference the top-level enrichment by default. This behavior helps maintain a standard reference point, but if you need only "main_body_text, " you may need to reselect it after each regeneration.
Sources: Deep Dive: Clay Scrapers [Limitless Research] - Clay University Transform: Clean, Structure, and Segment Data with AI - Clay University Intro to AI Snippets (Copywriting) [Clay 101] - Clay University
Richard X.
·
·
^ oof... Clay S. do we think we can stop the auto-defaulting if this is case? It's a lot of manual work telling the AI prompt to reference just the "main_body_content" each time after a regeneration. Lmk if I'm misunderstanding or if there's a better way to reference!
Channeled
APP
·
·
The documentation doesn't cover specific settings to prevent Clay from auto-defaulting to enrichment-level fields when regenerating AI prompts. The available guidance focuses on verifying that dynamic prompt inputs are correctly linked to appropriate columns after using the AI help tool. When configuring AI enrichments, you can verify that dynamic prompt inputs are correctly linked to the appropriate columns - this step is crucial when you have multiple columns connected to the same endpoint. Are you experiencing this auto-defaulting behavior specifically when using the "Generate with prompt" feature in the AI help tool, or does it happen with other types of AI prompt regeneration as well?
Sources: Referencing dynamic data from other columns | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University
Richard X.
·
·
yes, "Generate with prompt"
Channeled
APP
·
·
The documentation doesn't include specific settings to prevent Clay from auto-defaulting to enrichment-level fields when using "Generate with prompt." The available guidance focuses on verifying that dynamic prompt inputs are correctly linked to appropriate columns after using the AI help tool. When configuring AI enrichments with "Generate with prompt, " you need to verify that dynamic prompt inputs are correctly linked to the appropriate columns - this step is crucial when you have multiple columns connected to the same endpoint. Are you finding that the auto-defaulting happens immediately after clicking "Generate with prompt, " or does it occur when you make other changes to the prompt configuration?
Sources: Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University Writing AI prompts in Clay | Documentation | Clay University
Piotr M.
·
·
Richard X. What can really help is a function that generates prompts with AI. It uses the SPICED framework and also produces results automatically, without any additional parameters if you don’t want them.
LuisArturo
·
·
Hey there Richard sorry for the delay here, taking a look at this 1. The difference here is that by entering in the "Research3" all of the values inside of the cell will be used here vs just the single value that was selected in "main_body_text" 2. If you wish to have answers that would cause the AI to not hallucinate as much then entering in the "main_body_text" by itself would be best here. 3. Could you share the table link here so that we can check.
Channeled
APP
·
·