How to Use the Advanced AI Enricher

Before you start

• Define a clear enrichment goal:
  • Define what you want the AI to generate or extract.
  • Example use cases:
    • Write personalized LinkedIn connection messages.
    • Summarize or classify customer reviews.
    • Translate profile or text data into other languages.
    • Categorize leads by industry, seniority, or location.
• AI Phantoms don’t extract or post anything on your behalf, they use the data you already have to generate AI-written messages, insights, or other text results (depending on the Phantom you’re using).
• Awareness of credit usage:
  • Lightweight models (GPT-3.5 Turbo, GPT-4o Mini) → lower cost, best for testing and simple enrichments.
  • Advanced models (GPT-4, GPT-4o) → higher quality, but can consume significantly more credits.

Step 1: Choose your GPT model

Choose the model that will process your enrichment:

• PhantomBuster AI credits:
  • gpt-4o-mini (PhantomBuster) – enhanced capabilities, moderate cost (recommended default)
  • gpt-3.5-turbo (PhantomBuster) – lowest cost
  • gpt-4 (PhantomBuster) – advanced, higher cost
• Your own OpenAI API key:
  • gpt-3.5-turbo / gpt-3.5-turbo-16k
  • gpt-4, gpt-4-turbo, gpt-4o

Step 2: Provide your data

1. Choose the Enrichment Mode:
   Select how you want your data processed:
   • Row by row (Individual Rows):
     Each row is processed separately, one API call per row.
     → Best for personalized enrichments, translations, or per-lead classification.
   • All at once (Bulk):
     The entire dataset is processed in one API call.
     → Best for summarization, sentiment analysis, or trend detection across many rows.
2. Select a prompt template or write your own:
   • Use one of the pre-written prompt templates:
     • Personalized LinkedIn connection message
     • Content analysis and sentiment detection
     • Text translation
   • Edit a template directly:
     • Once selected, the full prompt text appears in the editor.
     • You can adapt the instructions to your specific needs (e.g., change the tone, add formatting rules, request multiple outputs).

   • Or create a custom prompt tailored to your needs (Prompt examples).

     • The more specific your instructions, the better the output (define the task, specify the format, etc.).

     

3. Choose how to supply the data to enrich:
   • A URL:
     • Provide a Google Sheet with the data to enrich (make sure it’s shared with “Anyone with the link”).
     • Upload a CSV file with the data to enrich (make sure it’s publicly accessible, and note that CSV upload is only available on paid plans).
   • My Phantoms:
     Use results from another Phantom as input.
4. Specify columns to feed to GPT:
   • Tell the Phantom which columns from your spreadsheet the AI should use as input:
     • This step is required. If you don’t enter column names, the Phantom won’t run.
     • Type the column headers exactly as they appear in your spreadsheet (case-sensitive).
     • Enter one column name per line.

AI Phantoms require all selected fields to be present in your input data to function correctly.
For example, if your prompt references First name and Last name but some leads don’t include those fields, the Phantom will return this warning:
“All input data has been filtered out. Ensure it contains the properties specified in the property list.”
→ If some data is missing, first run a LinkedIn Profile Scraper to enrich your leads before using this AI Phantom.

Step 3: Adjust behavior

You can fine-tune your Phantom run:

1. Temperature (min 0, default = 1, max 2):
   The Temperature setting controls how focused or creative the AI’s responses are:
   • Lower values → more accurate, reliable, and consistent results (ideal for structured or factual tasks).
   • Higher values → more creative and varied responses, though results may occasionally be less relevant.
2. Rows per launch:
   • Limit how many rows are processed per run.
   • Default = empty, the Phantom processes all rows.

1. Name your results file (optional)

   • You can customize the file name.

     If you rename the file between launches, the Phantom will create a new results file and start processing inputs from scratch.

2. Fields to keep (optional)
   • Export only the fields you care about by listing them (comma-separated).
     Example: profileUrl, job, location.
   • This generates a second CSV file with just those fields.
   • Learn more in our guide: Customize how your Phantom results files are saved.

Step 4: Select launch frequency

Choose how often the Phantom should run:

1. Launch manually: Start the Phantom yourself whenever you need.
2. Launch once at a specific time: schedule a one-time run at a set date and time.
3. Launch repeatedly: schedule regular runs (e.g. once per day, several times during working hours).
4. Launch after another Phantom: chain automations together so this Phantom starts right after another finishes.
5. Advanced scheduling: customize the exact minutes, hours, days, or months when the Phantom should run.

→ For a complete walkthrough of scheduling options, see our guide to scheduling Phantoms automatically.

Step 5 (Optional): Advanced settings

Advanced settings are available if you want to fine-tune how your Phantom runs, but by default they’re already optimized for most use cases.

We recommend leaving them as they are unless a guide specifically instructs you to change something.

→ For a detailed overview of all advanced options (like execution limits, retries, email notifications, proxies, webhooks, and file management), see our Advanced settings guide.

Launch and results

When you’re ready:

1. Click Launch to start your Phantom.
2. Once it finishes, open the Results tab in the Phantom console.
3. Download your results as a CSV or JSON file.

Outputs are added as new columns in your results file.

• Row by row mode: each row gets its enrichment in a new column.
• Bulk mode: results may be structured (e.g., JSON) in a single column or cell.

→ To learn how to export your data to Google Sheets, integrate with other tools, or reuse it in more automations, check our Access and Export your Phantom Results guide.

Export and input limits on the Free plan
If you’re on the Free plan or Free trial, some features are limited:
- CSV exports include only the first 10 rows of results.
- CSV download links (for dynamic viewing in Google Sheets or integrations) are not available.
- JSON exports are not available.
- CSV upload as an input method is not supported.
To unlock all features, you’ll need to upgrade to a paid plan.

What you give (Input) and What you get (Output)

This section gives you a detailed breakdown of everything you need to provide to run this Phantom, and everything you’ll receive once it completes.

What you give (Input)

What you get (Output)

Tips and troubleshooting

Common pitfalls

• Using a private spreadsheet (make sure it’s set to “Anyone with the link”).
• Wrong enrichment mode: don’t use Bulk for per-lead outputs, and don’t use Row by row for trend analysis.

If you run into errors or account issues

• Check how to troubleshoot your phantom using Logs.
• Browse the Fix Issues & Troubleshoot Errors section for solutions to common problems.
• Review our Automation Rate Limits by Platform guide.
• Check our Best Practices for Social Media Automation guide.

Suggested automations

After enriching your data with AI, you can:

• Generate personalized LinkedIn messages → with the AI LinkedIn Message Writer.
• Send outreach at scale → with LinkedIn Outreach or LinkedIn Message Sender.
• Further enrich leads → with LinkedIn Profile Scraper or LinkedIn Company Scraper.