Facebook Page Validator & Brand Slug Checker

Check a list of Facebook slugs and get back, for each one, whether it resolves to a real page — and whether that page actually belongs to the brand you expected.

Facebook answers HTTP 200 for pages that do not exist. A slug you invented, a slug from a stale CRM export and a slug belonging to a real company all come back 200 OK. Any checker built on status codes marks all three as valid. This actor distinguishes them.

The problem it solves

Two different failures hide in a list of Facebook URLs, and both look fine from the outside:

The slug is dead. facebook.com/some-company-that-rebranded returns 200 with an empty shell. Your outreach tool records a valid link, your team clicks it in three weeks and finds nothing.

The slug is alive but belongs to somebody else. This is the one that costs you. facebook.com/bukalapak is not the Indonesian marketplace — it is an app page called Eoddata beeip.com with 124 followers. Guessing brand slugs from company names produces these constantly, and an existence check waves every one of them through.

Give this actor an expected brand per slug and it reports EXACT, STRONG, WEAK or MISMATCH against the page's real name.

What you get per slug

Field	What it tells you
exists / status	Does the slug resolve to a page at all
name / pageId	Who the page actually is
brandMatch / brandMatchScore	Whether that is who you expected
followers / followersLabel	Audience size, plus the label Facebook rendered
primaryCategory / categories	How Facebook classifies the page
phones / websites / emails / addresses	Public contact details from the about tab
payloadTier / postsAreTrustworthy	How complete Facebook's answer was — see below

Quick start

Existence only:

{ "slugs": ["nike", "starbucks", "some-slug-you-are-unsure-about"] }

Existence plus brand verification:

{
  "pages": [
    { "slug": "bukalapak", "expectedBrand": "Bukalapak" },
    { "slug": "tokopedia", "expectedBrand": "Tokopedia" }
  ],
  "onlyBrandMismatch": true
}

nike, facebook.com/nike and https://www.facebook.com/nike/ all reduce to the same slug, so you can paste URLs straight from a spreadsheet.

Cleaning a list

Set onlyNotFound: true to get just the dead slugs — the rows to delete. Set onlyBrandMismatch: true to get just the ones pointing at the wrong company — the rows to re-research. The RUN_SUMMARY record carries the totals either way, including mismatchedSlugs.

The payload tier, and why it is on every record

Facebook returns one of three response sizes, and the boundaries were measured rather than assumed (20 pages × 3 repeats, deterministic):

Tier	Size	Meaning
HOLLOW	~330 KB	No such page. A known-fake control slug returns exactly this.
REDUCED	~0.7–1.4 MB	The page exists and the header is present, but Facebook withheld the post feed.
FULL	≥1.5 MB	Header plus roughly ten posts.

REDUCED is the tier that misleads people. On it, postsInPayload is 0 for pages that post every day — NASA came back reduced on three consecutive fetches, Coca-Cola carried two posts. It is sticky per page, not a random throttle, so retrying does not clear it.

That is why postsAreTrustworthy exists and why this actor does not sell you a posting-frequency metric. A zero post count on a REDUCED payload is an absence of evidence, not evidence of absence, and any tool that averages the two together will quietly tell you an active competitor has gone dormant.

Follower count, name, category and contact details are read correctly on REDUCED payloads, so those fields are unaffected.

Limits worth knowing

• Approximate follower counts. Facebook renders 39M, not 39,214,880. followers is the parsed integer; followersLabel preserves what was actually shown. For small pages the count is exact.
• Only public data. Whatever a logged-out visitor cannot see, this cannot see. Many pages publish no phone or email at all.
• No post feed, no engagement metrics. Paginating posts requires a doc_id that Facebook no longer ships in page HTML, so it is out of scope by design rather than by omission.
• Personal profiles are not pages. profile.php and the other reserved paths are rejected rather than guessed at.
• Brand matching is a signal, not a verdict. WEAK means "a human should look at this". Regional pages like Kopi Kenangan | South Jakarta score as STRONG against Kopi Kenangan, which is usually right but not always what you want.

Cost and proxy

One request per slug, no login, no cookies, no token handshake. Leave the proxy off — this surface applies no TLS fingerprint gate (all 16 profiles tested return 200) and no IP gate, so a proxy adds latency and cost for nothing.

Pages are large, up to ~20 MB each, and Facebook sends no content-length and ignores Range requests, so the full body must be downloaded to classify it. That is fast in practice (~1s for an 8 MB page) and the actor fetches concurrently; raise concurrency to go faster at the cost of memory.

Output shape

Every record carries _input, _source, _scrapedAt and recordType. recordType is PAGE for a checked slug, RUN_SUMMARY once per run, and ERROR for a slug that could not be fetched. Three dataset views are provided: Overview, Brand check and Contacts. Optional exportFormats also writes JSON, NDJSON, CSV or XLSX to the key-value store.

Development

pip install -r requirements.txt
python -m src            # needs an Apify runtime
python test_local.py     # runs the checker against live pages, no Apify needed

CRAWLING_METHOD.md documents where each field comes from and which markers the parser keys off.