General Osteopathic Council Register Scraper

UK Osteopath Register (GOsC) Scraper

Check UK osteopaths against the General Osteopathic Council (GOsC) statutory public register at osteopathy.org.uk. The actor reads the register's own server-rendered pages without a browser, login, cookie, API key or proxy, and returns structured records for patient checks, credentialing, directories and research.

What it returns

Each dataset item is one registrant:

• Identity: registrantId, registrationNumber, name, title, firstName, lastName, gender
• Registration: practisingStatus, statusCategory, isPractising, hasConditions, sanction, dateOfRegistration, dateOfRegistrationRaw
• Regulatory context: hasHearingNotes, hearingNotesUrl (the GOsC hearing-decision page linked from the register entry)
• Education: qualifications, placeOfTraining, qualificationDate, qualificationYear
• Contact: email, contactNumber
• Practices: practices[] with name, practiceUrl, address, postcode, openingHours, email, phone, website, homeVisits, disabledAccess, welshSpoken, plus optional latitude, longitude, distanceMiles
• Practice summary: primaryPracticeName, primaryPracticeTown, primaryPracticePostcode, practiceCount, and website (the first available practice website, which may belong to a secondary practice)
• Provenance: sourceUrl, searchQuery, source, recordType, scrapedAt

Empty values are omitted at every depth. Boolean practice facilities and derived status flags are retained even when false.

Three ways to query

search with no surname or location behaves like the old browse-all search. fullRegister is the explicit heavy-crawl mode and ignores surname, location and registration inputs while still honouring practice-feature and post-fetch filters.

Inputs

Saved tasks that still send postcode, name, registrationNumber or radiusMiles keep working: those legacy keys are mapped in code, not shown as inputs.

Examples

Find osteopaths with a surname near a postcode:

{
  "mode": "search",
  "surname": "Smith",
  "location": "SW1A 1AA",
  "radius": "10mi",
  "maxItems": 25
}

Check a mixed list of registration numbers:

{
  "mode": "byRegistrationNumbers",
  "registrationNumbers": ["1766", "8123", 10877, "99999999"],
  "maxItems": 10
}

Find regulatory cases:

{
  "mode": "search",
  "status": "Conditions of Practice",
  "maxItems": 50
}

Build a directory sample:

{
  "mode": "fullRegister",
  "disabilityAccess": true,
  "maxItems": 500
}

Statuses and regulatory flags

The register publishes six practising statuses. The actor keeps the original text in practisingStatus and adds a stable machine-readable layer:

An unrecognised future status is preserved in practisingStatus, mapped to statusCategory: other, and logged once. Regulatory entries carry hearingNotesUrl and hasHearingNotes; the linked PDF is not parsed.

Location and distance

When location is supplied in search mode, the actor geocodes the origin through postcodes.io (postcode, outcode, then place-name lookup). Practice postcodes are geocoded in batches of 100, cached for the run, and used to add latitude, longitude, distanceMiles on each practice and the lowest distance on the record. Distance is haversine miles from the origin. If geocoding fails, records are still returned without distance fields; the run never fails because of the geocoder.

How pagination works

The register shows 12 practice cards per page and states the total as results of N. The actor computes ceil(N / 12) pages, so the final unlinked page is included (the old actor missed it). It stops on maxItems, an empty page, a repeated page signature, no new registrant IDs, or the computed last page. It does not use the largest pagination link as the total. A full browse currently needs 481 pages; pages beyond the last repeat the last page.

Practice cards can list several practitioners, so one page can yield more than 12 registrant IDs. Records are deduplicated by registrantId. The 5,761 browse figure is practices, not people; the current browse resolves to 5,688 unique registrant IDs.

Data quality

• Placeholder practices such as Practice Details, Practice Details Pending and Not Listed are removed.
• Practices are deduplicated by normalised name, address and phone, so legitimate branches at different addresses are kept.
• Compact and oddly spaced postcodes are normalised; unparseable postcodes are omitted rather than guessed.
• Websites gain https:// when the source prints a bare domain; non-URL values are dropped.
• Not Listed, N/A, punctuation-only phone numbers and other sentinels are omitted.

Limitations

• The browse is practice-centric. Osteopaths with no practice listing, including some suspended registrants, do not appear in fullRegister; exact surname or registration lookup still finds them.
• status, firstName, qualificationContains, registeredAfter and registeredBefore are post-fetch filters. A search that uses only those filters pages through the register to find matches, so it can be slow; combine them with a surname or location where possible. A status-only Suspended search cannot find practice-less suspended records in the browse; use a surname or registration number for those.
• Removed or erased registrants are not returned by the public search at all.
• The register's robots.txt disallows /register-search/. The actor uses low concurrency, polite pacing and only the pages needed for the requested records; it is intended for verification and directory use, not bulk republication.
• Hearing-decision documents are linked, not downloaded or text-extracted.
• Overseas practices often have no UK postcode, so distance and coordinates are omitted for them. Town and county extraction from overseas addresses is best-effort and may return a region or postal-code fragment when the source address has no clear town.
• The full register is a large crawl: request maxItems deliberately and allow a generous run timeout for thousands of rows.

Performance and reliability

The actor uses HTTP/2 server-rendered requests at 1024 MB, with detail concurrency 6 and 0.2–0.4 s pacing. It retries 403/429/5xx responses with exponential backoff, skips records that cannot be parsed, and resumes safely after a platform migration without duplicate rows or overshooting maxItems. It needs no browser, credentials or proxy.

FAQ

Is this the official register? Yes. Every record comes from the General Osteopathic Council's own public register search at osteopathy.org.uk/register-search/, the same register patients use to verify an osteopath.

Why does a partial surname return nothing? The register's surname search is exact and case-insensitive. Smith matches Smith, but Smit does not. Use the full surname, or combine a first-name filter with location or practice features.

Why is the browse total 5,761 but the full register smaller? The published 5,761 is the number of practice cards. Each card can list several practitioners, and many practitioners appear on more than one card. Deduplicating registrant IDs gives about 5,688 unique people today.

Can it find a suspended osteopath with no practice? Yes, through exact surname or registration-number lookup. Such a person is absent from the practice-centric browse, so fullRegister cannot enumerate them.

Do I need a login, cookie, API key or proxy? No. The register is public and the actor makes anonymous requests. Nothing is stored or sent to a third party except optional postcode geocoding through postcodes.io.

What do the status flags mean? isPractising is true for Fully Registered, Practising Overseas, Conditions of Practice and Undertaking, and false for Non Practising and Suspended. hasConditions is true for Conditions of Practice and Undertaking. sanction names the regulatory status for Suspended, Conditions of Practice and Undertaking.

Where do hearing notes come from? Regulatory entries publish a View hearing notes link beside the status. The actor stores its absolute URL in hearingNotesUrl; it does not parse the linked decision document.

Why do some practices have no postcode or distance? Overseas practices and some incomplete addresses do not contain a UK postcode. The actor keeps the address text and omits unparseable postcode, coordinate and distance fields rather than inventing values.

How long does a full-register run take? The register has 481 pages of practice cards and about 5,688 unique registrants. Detail enrichment is concurrency-limited and paced politely, so a complete crawl takes minutes and should be run with a deliberate maxItems and an adequate timeout.