Extract Phone Numbers from Any Text: 5 Methods Compared

You have a wall of text — chat logs, customer notes, scraped HTML — and you need every phone number in it. Here are five practical methods.

Method 1: Online extractor (zero install)

Paste your text into our number extractor, enable Include separators, and get a list of every digit cluster including phone-like patterns.

Best for: one-off extractions, sensitive data, anyone non-technical.

Method 2: Custom regex

# US/international phone numbers (loose)
\+?\d{1,3}?[-.\s]?\(?\d{3}\)?[-.\s]?\d{3}[-.\s]?\d{4}

# E.164 format (strict)
\+\d{8,15}

Run with grep -oE on the command line. Faster on huge files than any online tool.

Method 3: libphonenumber (Google’s library)

Google’s libphonenumber is the gold standard. It validates numbers against country-specific rules.

from phonenumbers import PhoneNumberMatcher
text = "Call me at +1 (415) 555-0123 or 020 7946 0958"
for m in PhoneNumberMatcher(text, "US"):
    print(m.raw_string)

Best when: you need validated phone numbers, especially across regions.

Method 4: NLP entity recognition (spaCy, Hugging Face)

For unstructured text where phone numbers are written naturally, an NLP pipeline can pick them out by context. Slower and more setup.

Method 5: LLM (Claude, GPT)

Send the text to a model and ask for a list. Highest accuracy. Costs money, sends your data to a third party, slowest.

Comparison table

The pragmatic choice

For 80% of cases, the online extractor or a regex is enough. Reach for libphonenumber when you need to validate the numbers. LLMs are a last resort.