AI Analysis with LLM Fields =========================== LLM prompt fields let you use AI to analyze every page on your site systematically. You define questions (prompts) and Content Chimera sends each page's content to an AI model, stores the answers, and makes them available as chartable data fields. This turns unstructured page content into structured, comparable data — so you can chart, filter, and report on things like content quality, topic classification, or audience suitability across your entire site. .. contents:: On this page :local: :depth: 2 What Are Fieldsets and Fields? ------------------------------ LLM analysis is organized into two levels: **fieldsets** and **fields**. A **fieldset** is a group of related prompts that are run together. Think of it as a scoring rubric or a classification scheme. For example, you might have a fieldset called "Content Classification" or "E-E-A-T Scoring." A **field** is an individual question within a fieldset. Each field defines a specific prompt that will be sent to the AI model along with each page's content. For example, within an "E-E-A-T Scoring" fieldset, you might have fields for "Expertise Score," "Authoritativeness Score," and so on. Fields have a **result type** that controls what kind of answer the AI returns: - **String** — A text answer (e.g., "Blog Post," "Product Page," "FAQ") - **Number** — A numeric value (e.g., a score from 1 to 5) - **Structured** — A JSON object for more complex responses Fields can also have **constrained values** — a predefined list of allowed answers. This works like a dropdown menu: you tell the AI it must choose from specific options (like "Blog Post," "Product Page," "Landing Page," "Support Article"). Constraining values makes the results more consistent and easier to chart. Each field can be **required** or optional, and can allow **single or multiple responses** (for example, a page might belong to more than one topic category). Templates --------- Content Chimera includes built-in **fieldset templates** — pre-configured sets of prompts designed for common analysis tasks. Templates save you from writing prompts from scratch. Available templates include: - **E-E-A-T Scoring** — Rates each page on Google's content quality dimensions (see :ref:`eeat-scoring` below) - **Content Classification** — Categorizes pages by type and purpose - **AI Readiness** — Evaluates how well content is structured for AI consumption To use a template, copy it to your extent. The copy becomes your own version that you can customize — change prompts, add or remove fields, or adjust the allowed values. .. admonition:: Web UI :class: tip Go to **LLM Fieldsets** from the extent menu. You'll see a list of available templates alongside any fieldsets you've already created. Click **Copy** on a template to add it to your extent. .. admonition:: MCP :class: tip Ask your AI assistant to show you the available templates and copy one to your project: *"What LLM fieldset templates are available?"* *"Copy the E-E-A-T Scoring template to this extent"* .. raw:: html
Tool reference

Tool: llm-fieldsets

action_type: "list"    — list templates
      action_type: "copy"    — copy a template
      source_fieldset_id: 56 — which template to copy
      extent_id: 1234
Creating Custom Fields ---------------------- If the templates don't cover what you need, you can create your own fieldsets and fields from scratch. When writing prompts, keep these principles in mind: - **Be specific.** "Rate this page's expertise on a scale of 1-5" works better than "How expert is this?" - **Define the scale.** If you want a number, explain what each value means (e.g., "1 = no expertise demonstrated, 5 = deep subject-matter expertise with citations"). - **Use constrained values when possible.** Giving the AI a fixed list of options produces more consistent, chartable results. - **Test before bulk-running.** Always test your prompts on a few representative pages before running them across the full site (see the next section). .. admonition:: Web UI :class: tip 1. Go to **LLM Fieldsets** and click **Create Fieldset** 2. Give it a name and optional instructions (context provided to the AI for all fields in this fieldset) 3. Add fields one at a time — each with a name, prompt, result type, and optional value constraints 4. Choose the AI model to use and whether to send the full HTML or simplified text .. admonition:: MCP :class: tip Ask your AI assistant to create a fieldset and add fields to it. Describe what you want each field to do: *"Create an LLM fieldset called Content Classification"* *"Add a field called Content Type that classifies each page as Blog Post, Product Page, Landing Page, Support Article, or Other"* Your AI assistant will handle the technical details — setting up the prompt, result type, and allowed values. .. raw:: html
Tool reference

Tools: llm-fieldsets and llm-fields

Result types: 1 = String, 2 = Number, 3 = Structured

Testing Before Running ----------------------- Before running your fieldset across every page on the site, test it against a single URL. This lets you see exactly what the AI returns and refine your prompts without waiting for a full bulk run. .. admonition:: Web UI :class: tip From the **LLM Fieldsets** page, open your fieldset and click the **Test** button. Enter a URL and Content Chimera will fetch the page, run all fields against it, and show you the results. .. admonition:: MCP :class: tip Ask your AI assistant to test your fieldset against a sample page: *"Test this fieldset against https://example.com/blog/sample-post"* The response shows the AI's answer for each field, so you can verify the prompts produce useful results before committing to a bulk run. .. raw:: html
Tool reference

Tool: llm-fieldsets with action_type: "test_example_url"

Running Bulk Summarization -------------------------- Once your fieldsets are configured and tested, you can run them across all pages in your crawl. This is called **bulk summarization** — Content Chimera sends every page's content to the AI model and stores the results. .. important:: Bulk summarization sends each page to an AI model, which has an associated cost. You will always be asked to confirm before the run starts. The confirmation tells you how many pages will be processed so you can estimate the cost. Before running, make sure your fieldsets are attached to a **summary definition** — this tells Content Chimera which fieldsets to include in the bulk run. You can have multiple fieldsets active at once. .. admonition:: Web UI :class: tip 1. Confirm your fieldsets are listed under the active **Summary Definition** in your extent settings 2. Go to **My History** and choose **Run Summarization** from the actions menu 3. Review the confirmation prompt showing the number of pages and fieldsets 4. Click **Confirm** to start the run .. admonition:: MCP :class: tip Ask your AI assistant to run bulk summarization: *"Run LLM summarization for this extent"* Your AI assistant will check which fieldsets are configured, show you a confirmation with the estimated scope, and proceed only after you approve. .. raw:: html
Tool reference

Tools: summary-definition (check configuration) and run-focused-pipeline with pipeline: "summarize"

Using the Results ----------------- After bulk summarization completes, every field value from every page becomes a column in your **flattened table**. This means you can use them just like any other data field in Content Chimera: - **Chart distributions** — How many pages scored 4 or 5 on Expertise? What percentage are classified as "Blog Post"? - **Filter and sort** — Show only pages with a low Trustworthiness score, or focus on pages the AI classified as "Landing Page." - **Use in rules** — Create rules that act on AI-generated fields (e.g., flag all pages with an Expertise score below 3). - **Include in reports** — Add charts based on LLM fields to your Chimera reports. - **Cross-reference with other data** — Combine AI scores with analytics data, crawl depth, or word count for deeper analysis. .. admonition:: Chimera Chat :class: tip Once summarization is complete, you can ask questions about the results: - *"What's the average Expertise score across the site?"* - *"Show me a chart of Content Type distribution"* - *"Which pages scored below 3 on Trustworthiness?"* .. admonition:: MCP :class: tip Ask your AI assistant to query or chart the results, just as you would with any other data field. For example: *"Show me a chart of Content Type distribution"* *"What fields are available from the LLM analysis?"* .. _eeat-scoring: E-E-A-T Scoring ---------------- E-E-A-T stands for **Experience, Expertise, Authoritativeness, and Trustworthiness**. It is the framework Google uses to evaluate content quality. Pages that score well on these dimensions tend to perform better in search results, especially for topics where accuracy matters (health, finance, legal, etc.). The four dimensions: - **Experience** — Does the content demonstrate first-hand experience with the topic? - **Expertise** — Does the author show deep knowledge and competence? - **Authoritativeness** — Is the author or site recognized as a go-to source on this topic? - **Trustworthiness** — Is the content accurate, transparent, and honest? Content Chimera includes a **built-in E-E-A-T fieldset template** that scores each page on all four dimensions using a 1-5 scale. The AI reads each page and assigns scores based on signals in the content — things like cited sources, author credentials, specific examples, and balanced presentation. **How to use it:** 1. Copy the E-E-A-T template to your extent (see the Templates section above) 2. Optionally customize the prompts or scale 3. Test against a few representative pages 4. Run bulk summarization 5. Chart the results to identify content quality gaps — for example, pages with high Expertise but low Trustworthiness might need better source citations E-E-A-T scoring is particularly useful for: - **Content audits** — Finding pages that fall below a quality threshold - **Prioritizing improvements** — Focusing editorial effort where scores are lowest - **Client reporting** — Showing stakeholders a data-driven view of content quality - **AI Readiness assessments** — E-E-A-T scores are one component of the broader AI Readiness pipeline