As of 13 August 2026, AI can only partly remove spam leads from your sales pipeline.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your CRM and export the leads you want to screen, including the lead ID, contact details, source, message, status, owner, dates and notes.
   2. Write down your business definition of spam, including examples of genuine leads that must not be removed and whether records should be quarantined rather than deleted.
   3. Remove unnecessary personal data from the export, keep the lead IDs, and paste the remaining table into the prompt with the business rules filled in.
   4. Ask the chatbot to classify each record as definite spam, likely spam, needs human review, or probably genuine, with evidence and a recommended action.
   5. Compare every definite-spam recommendation with the original CRM record, email domain, website and source, then manually inspect all likely-spam and human-review records.
   6. Ask the chatbot to revise the rules using the confirmed examples, and test those rules on a separate sample of leads before applying them to the full export.
   7. Create a reversible CRM view or quarantine status for approved records, retain the lead IDs and reasons, and only then archive or delete records under your normal CRM process.

   Prompt

   I need to identify likely spam leads in a UK sales pipeline without deleting genuine prospects. I will paste a CSV or table of lead records below, with these fields where available: [lead ID], [name], [company], [work email], [website], [source], [message], [created date], [last activity], [owner], [status], and [notes].
   
   Use these business rules: [describe what counts as spam for this business]. Treat these as warning signs, not automatic proof: disposable or malformed email addresses, impossible or irrelevant company details, repeated submissions, obviously promotional messages, empty or nonsensical enquiries, and records matching known spam patterns. Do not infer spam from a person's name, nationality, location, age, gender or any other protected or sensitive characteristic.
   
   Return a table with the lead ID, classification of definite spam, likely spam, needs human review, or probably genuine, the evidence for the classification, and a recommended action of quarantine, keep, or investigate. Do not invent facts, enrich missing details, or recommend permanent deletion. Flag every uncertain case for human review. End with proposed rules that I can test on a sample before applying them to the rest of the pipeline, and list the records where one false positive would be especially costly.
   
   Here is the data:
   [PASTE DATA HERE]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

3. Hand it to a person

   The distant third

   A person who owns the outcome does this end to end, worth it when the failure is dear.