personalized first lines for cold email, written from each prospect's actual website

this AI icebreaker generator visits each lead's real website (homepage, about, services, latest blog posts), pulls something specific and citable, and writes cold email opening lines that prove you actually looked. not a bio rewriter, not a template spinner: a first line writer that reads the actual site.

generic first lines kill reply rates. "i hope this email finds you well" gets deleted on sight, and most "personalization" tools just reword the prospect's LinkedIn headline. specific beats generic, every time.

📊 what the output looks like

lead	icebreaker	confidence
roofing company, Tampa	saw you guys just wrapped that 40-unit re-roof in Brandon, that's a serious project for a 6-man crew.	high
landscaping company, FL	your blog post on florida-friendly plants actually answered something i'd been wondering about st augustine grass.	high
dev tool startup	just read the gradcracker debugging story on your blog, three days of ai hallucinating solutions sounds painfully familiar.	high
site unreachable	figured i'd reach out to Dead Site Inc directly instead of going through a contact form	low

every row keeps all your input columns and adds icebreaker_1 to icebreaker_3, fact_used (the signal it cites), confidence (high / medium / low), signal_source (homepage / about / services / blog / fallback) and detected_language when you use auto language.

⚡ quick start

paste leads as JSON, or chain another actor's run with datasetId. each lead needs a website field (websiteUrl, website, url or domain all work). companyName and contactName are recommended but optional.

{
    "leads": [
        { "websiteUrl": "https://tampabayroofers.com", "companyName": "Tampa Bay Roofers", "contactName": "Mike" }
    ],
    "senderContext": "i run a web design agency for home service contractors",
    "tone": "casual",
    "language": "en",
    "variantsPerLead": 2
}

senderContext is used to angle the line toward relevance. it is never pitched: the icebreaker only proves you looked.

📥 export to instantly, smartlead or lemlist

download the dataset as CSV and import directly. no column edits needed.

instantly

1. contacts, import, CSV
2. map icebreaker_1 to a custom variable named icebreaker
3. use {{icebreaker}} as the first line of your email body

smartlead

1. lead list, import CSV
2. map icebreaker_1 to a custom field (e.g. first_line)
3. reference it as {{first_line}} in your sequence

lemlist

1. campaign, leads, import CSV
2. map icebreaker_1 to a custom variable
3. insert it with the variable picker in your template

tip: filter the CSV to confidence != low for your A-list sends, and keep the low rows for a separate generic-angle campaign.

🔗 chain it with your lead source

built to sit at the end of a lead-gen pipeline. run any of these, then pass the run's default dataset id as datasetId:

• google maps lead generator or the pro version
• website email extractor
• tripadvisor lead generator
• email verifier + enricher

emails, phones and every other column your scraper found ride along into the final CSV.

🚫 why these lines don't sound AI-generated

most AI first line writers produce openers that scream automation: "i was impressed by your commitment to delivering seamless solutions". this actor enforces a strict style contract:

• one specific fact per line. a recent project beats a service list, a blog post title beats "i saw your website"
• banned words: impressive, amazing, streamline, leverage, seamless, robust, elevate, unlock, empower, game-changer
• banned patterns: "i hope this email finds you well", "i came across your website", "it's not just X, it's Y", exclamation points, em dashes
• lines that slip through get caught by a validation pass and regenerated

🌎 cold email opening lines in spanish

set language to es for spanish first lines, or auto to detect each prospect's website language and match it. a lead list mixing miami contractors and mexico city agencies comes back with each icebreaker in the right language, marked in the detected_language column. spanish-language cold email is badly underserved by first line tools, and this is built for it: the fallback templates, the style rules and the signal extraction all work natively in spanish.

🛡️ what happens with bad leads

runs never fail because of a bad lead:

• dead site, timeout or blocked request: the lead gets a fallback line from its company name, marked confidence: low
• linktree or beacons page instead of a real site: the actor follows the main link and crawls the real website behind it
• empty single-page app with no readable text: treated as unreachable, fallback line
• the run summary counts every fallback by reason so you can judge list quality at a glance

❓ faq

how is this different from pasting leads into chatgpt? scale and grounding. this visits every website, extracts real signals (projects, blog posts, certifications, years in business, service area) and cites them. no hallucinated compliments, no copy-paste loop, and the output lands as a CSV ready for your sequencer.

what does a good cold email first line look like? specific, casual, and about them. "saw you guys just wrapped that 40-unit re-roof in Brandon" reads like a person typed it after visiting the site. that is the target, and the style rules exist to protect it.

how many leads can i process per run? default cap is 500 per run, configurable with maxLeads. a 500-lead run typically finishes in a few minutes.

does it work with lead lists from other scrapers? yes. field names are normalized automatically (websiteUrl, website, url, domain, and companyName, company, name), so output from most scrapers chains directly via datasetId without remapping.

what if a lead has no website at all? it still gets a usable fallback line built from the company name, marked confidence: low and signal_source: fallback so you can filter it.

💸 pricing

pay per lead processed. a 1,000-lead list costs $8, which is 25-60x cheaper than human first line writers ($0.10-0.50 per line) and a fraction of what list agencies charge ($200-500 per 1k leads). the apify free tier covers a ~50 lead test run so you can check the quality on your own list before paying.