AI Personalization

AI Personalization automatically rewrites your pitch per website using the target site’s content as context. Instead of every recipient getting the exact same paragraph, each one gets a personalized version that references their business, industry, or offering.

Why it matters. Manual, personalized outreach converts 3-10x better than generic mass messages. AI Personalization gets you closer to hand-written quality while keeping the throughput of automation.

What the AI does. For each URL in a batch:

1. Reads the visible content on the target page (usually the homepage).
2. Extracts the site’s name, industry, and value proposition.
3. Rewrites your pitch’s placeholders (like {sell}) with a personalized angle referencing that specific business.

You control what the AI does via your agent’s pitch template and prompt.

1. Go to Agents in your dashboard.
2. Edit the agent you want to enable personalization for.
3. Scroll to AI Personalization.
4. Toggle Enable.
5. Write your prompt (or start from a template).
6. Add placeholders like {sell} to your pitch text where the AI should insert personalized content.
7. Save.

Test on a small batch (5-10 URLs) before enabling for a large run. See running campaigns for how to start a small test.

Placeholders mark spots in your pitch where AI-generated text should be inserted. Written in curly braces: {sell}, {value}, {intro}, etc.

How it works. When you write “I’d love to help you {sell}”, the AI reads the target site, decides what makes sense for that specific business, and replaces {sell} with something like “scale your Shopify boutique with paid ads” or “reach more B2B decision-makers”.

Placeholder rules:

• Names are free-form; use whatever makes sense.
• The AI resolves each placeholder based on context in your prompt.
• Multiple placeholders in one pitch are all resolved together in one AI call.

Cost: one AI call per URL (regardless of placeholder count), billed from your AI wallet balance.

The prompt is your instructions to the AI. Better prompts produce better personalization.

Good prompt structure:

1. State who you are. “You are writing for [your brand], a [category] tool for [audience].”
2. State the goal. “Rewrite the {sell} placeholder as one specific benefit the target business would care about.”
3. State the tone. “Casual, concise, one sentence maximum.”
4. Give an example. Concrete before/after transformations align the AI more than any instruction alone.

Testing prompts. Before running on a big batch, use the Preview feature in your agent to see how your prompt handles 5-10 sample sites. Iterate on the prompt, then run for real.

See agents and profiles for where prompts fit into your agent config.

AI wallet is a dollar balance used to pay for AI Personalization calls. Separate from your website credits.

1. Why separate. Website credits pay for form submissions (fixed cost per submission regardless of pitch complexity). AI wallet pays for language model tokens (variable cost per personalization call depending on prompt length + response length). Keeping them separate lets you use ContactID without AI, or use AI heavily without touching your website credits.

2. How it’s charged. Every AI Personalization call deducts a fraction of a cent based on tokens used. Typical costs:

• Simple prompt + short response: ~$0.002 per URL
• Complex prompt + longer response: ~$0.008 per URL

See the “Estimating AI Cost” article below for a more detailed breakdown.

3. Top up any time. Buy AI wallet dollars from Billing > AI Wallet in packages of $5, $10, $25, $50, $100. See plans and pricing.

Balance is shown in two places:

• Extension side panel. Top of the panel next to your website credits.
• Dashboard. Overview page and Billing > AI Wallet.

To top up:

1. Go to Billing > AI Wallet.
2. Click Top Up.
3. Pick a dollar amount.
4. Confirm on Stripe checkout.

Balance updates immediately after successful payment. You can also enable auto-top-up so it never runs out mid-batch.

ContactID routes AI Personalization requests through a pool of language models from three providers with automatic failover for reliability.

1. The three providers.

• OpenAI (GPT family)
• Anthropic (Claude family)
• Google (Gemini family)

2. Primary model. The default primary is a mid-tier general-purpose model, currently GPT-4o-mini. Chosen for:

• Speed. Under a second per request in most cases, keeps batch throughput high.
• Cost. Fraction-of-a-cent per call, predictable batch AI costs.
• Short-form quality. Personalization prompts ask for one or two sentences, exactly what this class of model handles well.

3. Failover behavior. If the primary is unavailable (rate-limited, temporarily down), requests automatically retry against a secondary from a different provider. This means your batches keep running even during outages at any single provider.

4. Why route across providers. Provider outages are the single biggest source of failed AI calls in production. Routing across OpenAI, Anthropic, and Google means one provider going down doesn’t stop your batch.

5. Checking the current model. The AI Personalization settings page shows the current primary model and failover list. This is subject to change as providers release new versions.

If AI wallet balance drops to zero mid-batch:

• Sending continues. AI Personalization is optional; if the wallet is empty, ContactID sends your unmodified base pitch instead.
• URL is still counted as a successful send if the form submits. Website credit is charged normally.
• You get a notification email so you know AI is disabled.

To resume personalization, top up AI wallet and the next URL in the batch uses AI again. See plans and pricing for top-up options.

Estimating cost before a big batch prevents surprises. A rough formula:

Cost per URL = (input tokens + output tokens) × per-token rate

• Input tokens: your prompt + target page text (typically 500-2,000 tokens per URL).
• Output tokens: the AI’s response (typically 30-100 tokens for a personalized snippet).
• Per-token rate: varies by model. GPT-4o-mini is ~$0.15 per million input tokens, ~$0.60 per million output tokens.

Rule of thumb: $0.002 to $0.008 per URL depending on prompt complexity. A 1,000-URL batch typically costs $2-8 in AI wallet.

To be exact: run a test batch of 20-50 URLs and check the AI wallet balance drop; multiply for your full batch size.

Two ways to test before committing to a large batch:

1. Preview feature in agent editor.

1. Open your agent in Dashboard > Agents.
2. Scroll to AI Personalization.
3. Click Preview.
4. Paste a target URL. AI runs your prompt and shows the personalized output.
5. Iterate on the prompt until output quality is right.

2. Small test batch.

1. Import a small list of 10-20 diverse target URLs.
2. Start a batch.
3. Review results in the Activity log with the sent-message field visible.
4. Fix prompt issues, then run the full batch.

Iterating on 10 test URLs is much cheaper than running a bad prompt on 1,000.