AI Personalization with AI-Powered Custom Fields

AI Custom Fields let you scale personalization in Pitchbox by using AI to extract or generate information from websites, pages and your other Custom Fields. This article covers connecting your AI integration, creating AI Custom Fields, using the AI Prompt Library, scraping options and output formats, and generating results in your campaign workflow.

Navigation

1. Connecting Your AI Integration
2. Where to Create AI Custom Fields
3. Creating an AI Custom Field
4. Writing Effective AI Prompts
5. Validating Your AI Prompt
6. Generating AI Custom Fields in Your Campaign Workflow
7. Monitoring AI Status and Processing Time

🔌 Connecting Your AI Integration

AI Custom Fields run on your own AI integration, so connect one before you start. Pitchbox supports OpenAI and Anthropic.

1. Go to the Integrations page in your Pitchbox settings.
2. Enter your API key for OpenAI or Anthropic, or choose Pitchbox AI.
3. Choose your AI model. More capable models usually give better results but cost more per request.

AI usage costs are billed by your AI provider, not by Pitchbox. Pitchbox charges no extra fees for AI Custom Fields, so you control spend through the model you choose and how many opportunities you personalize.

📍 Where to Create AI Custom Fields

You can create AI Custom Fields in three places. Each one opens the same setup form.

• Custom Fields page: open the main menu and select Custom Fields. Fields created here are available across your projects and campaigns, organized into groups.

• Campaign settings: open a campaign's settings and go to Custom Fields to create a field while you set up that campaign.
• Template editor: create a new AI Custom Field directly while writing an email template, so you can add it to your copy straight away.

When you add a new field, Pitchbox asks what type of field it is. Choose AI-Powered Field for fields that generate data using AI.

✨ Creating an AI Custom Field

The AI-Powered Custom Field form has five settings. Fill them in, then click Validate Prompt.

Name

Give the field a clear name. This is how it appears in your campaign workflow and as a variable in your templates.

Scraping URL

Choose which page the AI reads before it answers your prompt.

Use No Scraping when the field builds on information you already have, such as writing a subject line from other Custom Fields.

AI Prompt

Write the instructions for the AI. Type / or click Variables to insert other fields into your prompt.

To start from a proven example, click AI Prompt Library. The library is organized into two categories:

• Qualification: prompts that help you judge whether an opportunity is a good fit, such as Content Freshness in Days, Content Relevance Score, E-commerce Likelihood Scorer and E-E-A-T Assessment.
• Personalization: prompts that generate tailored content for your outreach.

Examples of our pre-made prompts are:

• Content Freshness (Days): How recently the page was published or last updated. Fresher content is often a better signal that the site is actively maintained and that links are less likely to be removed or neglected over time.
• Content Relevance Score: A measure of how closely the page’s content aligns with your target keywords or topic. Higher relevance increases the likelihood that a backlink appears natural and contextually appropriate.
• DMOZ Categorization: A legacy categorization system used to classify sites by topic. While DMOZ itself is no longer active, this categorization can still help infer the primary theme or industry focus of a domain.
• E-commerce Likelihood Scorer: Estimates whether a site is primarily transactional (selling products or services). This helps distinguish between editorial content sites and commercial pages, which may affect link placement value and intent.
• E-E-A-T Assessment: Evaluates signals related to Experience, Expertise, Authoritativeness, and Trustworthiness. Strong E-E-A-T indicators suggest higher editorial quality and a lower risk of spam or manipulative link behavior.
• Fuzzy Link Finder: Identifies links that may not be exact matches but are semantically related. This helps detect contextual relevance even when anchor text or phrasing varies naturally.
• Homepage Theme: Identifies the main topic or focus of the site’s homepage. This helps confirm whether the overall domain aligns with your target niche or industry.
• Image Optimization Score: Assesses how well images are optimized (alt text, file size, usage). While not a direct link metric, good image optimization often correlates with higher overall content quality and SEO maturity.
• Is Article?: Determines whether the page is a true editorial article rather than a category, product, or utility page. Editorial articles are typically better candidates for contextual backlinks.
• Is Listicle?: Flags whether the page is list-based (e.g., “Top 10…”). Listicles can be good link opportunities but are sometimes overused in paid placements, so this helps with risk assessment.
• Page Type: Classifies the page format (blog post, landing page, category page, etc.). This helps ensure backlinks are placed in appropriate, editorial contexts rather than purely commercial pages.
• Post Sitemap (source: Primary Sitemap): Indicates that the page appears in the site’s post sitemap, suggesting it’s a recognized piece of content rather than an orphaned or autogenerated page.
• Post URL (source: Post Sitemap): Confirms the page URL as listed in the sitemap, reinforcing that the page is intentionally published and maintained.
• Primary Sitemap (source: Robots.txt): Identifies the main sitemap declared by the site. A properly configured sitemap is a baseline signal of technical SEO health and legitimacy.
• Target Audience Identifier: Determines who the content is written for (e.g., consumers, professionals, local users). Backlinks are stronger when the linking page’s audience matches the target audience of the linked site.
• Top 3 Page Themes: Highlights the primary themes discussed on the page. This helps confirm topical alignment and avoids placements on pages that drift across unrelated subjects.
• Topic Authority Explained: Assesses how deeply and consistently the site covers a specific topic. Sites with demonstrated topical authority tend to provide more credible and durable backlinks.
• Topical Relevance: Measures how closely the page and domain align with the target topic overall. High topical relevance reduces the risk of links appearing forced or unnatural.
• Write-for-Us Detector: Identifies whether a site actively solicits guest posts. While not inherently negative, aggressive “write-for-us” behavior can indicate link-selling or lower editorial standards.

Pick a prompt, then edit it to fit your campaign.

Output Format

Choose the type of result you want back: Single-Line, Multi-Line, Number, Date, URL, Dropdown, Checkbox, Currency or On/Off Toggle. Match the format to how you will use the result. For example, use Number for a relevance score you want to filter by, or Multi-Line for a personalized paragraph.

Output Description

Describe exactly what a good result looks like, including length, tone and structure. For example: "Return a brief, personalized comment under 1000 characters with casual tone and specific praise. End after the value comment to allow for CTA insertion."

✍️ Writing Effective AI Prompts

The more specific your prompt, the more useful the result. Say exactly what to extract, where to look and what format to answer in.

• Yes/No answers: "Respond with 'Yes' if there is a 'Write for Us' link in the [footer] section, otherwise 'No'."
• Scores: "On a scale of 1 to 10, rate the content's relevance to [your topic], where 1 means not relevant and 10 means highly relevant."

Available Parameters

Reference parts of the scraped page in your prompt to target the AI more precisely.

When working with links, you can also reference:

✅ Validating Your AI Prompt

Click Validate Prompt to check your prompt before you save the field. Pitchbox scores it from 0% to 100% based on clarity, how well it uses the available parameters, and whether it fits what Pitchbox can do.

If the score is below average, you get specific suggestions, such as simplifying the wording or adding a parameter. Update the prompt and validate again until you are happy with the score.

🚀 Generating AI Custom Fields in Your Campaign Workflow

You generate AI Custom Fields for your opportunities inside the campaign workflow. Each run uses your connected AI integration and model, and usage is billed to your provider account.

AI Custom Fields are most useful at two stages:

• Inspection: run qualification fields to check whether opportunities fit your campaign, then filter out or delete the ones that don't.
• Personalization: run personalization fields to tailor each opportunity's outreach and improve response rates.

You can generate results in two ways:

• All AI Custom Fields: generate every AI Custom Field for all or selected opportunities.
• Specific AI Custom Fields: choose individual fields to generate. This is useful when you have refined a prompt and want to regenerate only that field.

📊 Monitoring AI Status and Processing Time

Generating AI Custom Fields usually takes a few seconds per opportunity. Bulk runs can take a minute or more, and Pitchbox marks each opportunity as it completes.

Track progress in the AI Status column. A red indicator means an error occurred for that opportunity. Click the indicator to see the error message and fix the issue.

You can view your AI Personalization Activity in the Campaign Card Mega Menu.

💬 Need more help? Email pitchbox@support.com or use the live chat and our team will be happy to help.