Contractor lead source cleanup before AI summaries.

AI summaries can make a contractor's lead list easier to scan, but they become shaky when every record says "website," "phone," or "other." Clean source labels help the owner see what happened before a draft reply or agency report is trusted.

Why vague source labels create bad summaries.

A summary that says "12 website leads need follow-up" is not very useful. It does not tell the owner whether those leads came from an emergency plumbing form, a roofing storm page, an HVAC no-cool landing page, an old estimate page, or a generic contact form. Each source has a different urgency and a different follow-up standard.

When the source is vague, AI may group unlike records together and make the queue look simpler than it really is. That can hide urgent demand, duplicate records, unowned callbacks, and agency campaign handoffs that need review. The cleanup goal is not fancy attribution. It is owner-visible context before action.

A useful source map for local service teams.

• Emergency form: leak, flood, no heat, no cooling, clogged drain, storm damage, or after-hours service demand.
• Quote form: normal estimate request, remodel inquiry, replacement quote, maintenance inquiry, or non-urgent project question.
• Phone call: inbound call, missed call, voicemail, booked call, no-answer callback, or call-back requested.
• Google Business Profile: call, message, direction request, website click, or local discovery action that should connect to service-area proof.
• Referral or repeat customer: relationship-based demand that may need a more personal reply than a cold web lead.
• Old estimate: stale quote, waiting-on-customer record, close-out reminder, or respectful check-in.
• Agency campaign: paid search, local SEO, landing page, email, or partner campaign source that needs reporting context.

The cleanup workflow.

1. Export or sample the last 30 days of leads from the CRM, forms, call tracking, inbox, and text system.
2. List every source label exactly as it appears, including messy versions like "web," "website form," "contact us," and "landing page."
3. Merge duplicates only when they truly mean the same thing. Do not merge emergency forms with ordinary estimate forms.
4. Add a plain-language owner note for labels that still need interpretation.
5. Run a small review with the Follow-Up Cleanup Checklist before using AI summaries for customer-facing follow-up.
6. Use the AI Reply Risk Checker when a summary becomes a draft reply.

Agency reporting benefit.

For agencies, clean source labels can make client conversations calmer. Instead of defending traffic quality from a blended lead list, the agency can show which demand came from paid search, local SEO, repeat customers, referral paths, emergency pages, or old estimate recovery. That does not prove performance by itself, but it helps separate marketing inputs from follow-up execution.

When a contractor is considering AI search or GEO work, the same source map also helps the agency choose better proof pages. A page about emergency plumbing calls needs different examples than a page about roofing storm intake or old estimate follow-up. The Agency Partners page outlines the safer partner path for this kind of cleanup support.

What to avoid.

Do not let AI invent a source when the record is unclear. Do not turn "unknown" into "organic search" because that sounds better. Do not use source cleanup to make a campaign look stronger than the evidence supports. And do not use a summary as proof that a lead was contacted unless the last human touch is visible.

A small proof sample before scaling.

Start with a small proof sample instead of a full CRM rebuild. Pick ten recent records: two emergency requests, two ordinary quote requests, two phone calls, two old estimates, one referral, and one agency campaign lead. For each record, write the visible source, the actual customer request, the owner, the last human touch, and the next step. If the record cannot answer those five points, mark it as source cleanup needed.

This small sample helps owners and agencies avoid arguing from averages. One record may show that the campaign source is fine but the callback owner is missing. Another may show that the lead source is unknown because the form tool and CRM are not passing the same field. Another may show that old estimate follow-up is being mixed with new demand. Those are different problems, and they deserve different fixes.

After the sample is clean, AI can be used more safely for internal summaries: "three urgent records need same-day review," "two quote requests are waiting on photos," or "one agency campaign lead has no owner." Keep those summaries internal until a human checks the source and follow-up facts. Customer-facing messages should still be reviewed for tone, consent, offer boundaries, and factual support.

Sources worth keeping nearby.

• Google Search Central: Creating helpful, reliable, people-first content
• FTC business guidance on advertising and marketing
• FTC Advertising FAQ's: A Guide for Small Business
• FTC CAN-SPAM Act compliance guide for business