Facebook Marketplace — Technical Analysis

Platform Overview

Facebook Marketplace is the largest informal rental listings source in Montreal. No official API exists for Marketplace data. Meta is aggressive about anti-scraping enforcement.

Data Available Without Login

Field	Available	Notes
Title	Yes	Listing title
Price	Yes	Rent amount
Location	Yes	City/neighborhood
Images	Yes	Thumbnails visible
Listing URL	Yes	Direct link
Description	Partial	Truncated on cards
Phone number	No	Behind login wall
Seller name	No	Behind login wall
Full description	No	Requires clicking into listing (auth modal)

Technical Architecture

Component	Details
Frontend	React SPA, heavy JS
Data loading	GraphQL (internal, tightly guarded)
CDN	Facebook's own infrastructure
Anti-bot	Extremely aggressive

Anti-Bot Measures

Measure	Status
JS challenges	Yes — dynamic content verification
Rate limiting	Yes — strict frequency controls
IP blocking	Yes — geographic + reputation filtering
Behavioral analysis	Yes — detects non-human patterns
Browser fingerprinting	Yes — webdriver detection, canvas, WebGL
Account suspension	Yes — logged-in scrapers get banned fast
DOM obfuscation	Yes — class names change frequently
CAPTCHA	Yes — triggered on suspicious behavior

Scraping Approaches

1. Playwright with stealth (best option)

• Launch with --disable-blink-features=AutomationControlled
• Inject anti-detection scripts (hide navigator.webdriver)
• Dismiss login modal (click outside / press Escape)
• Simulate infinite scroll to load more listings
• Parse listing cards from DOM
• 3-7 second delays between actions

Pros: Can access public data without login Cons: Fragile selectors, constant maintenance, slow, resource-heavy

2. Session cookie reuse

• Log in manually once in real browser
• Export cookies
• Load cookies into Playwright context
• Access more data (descriptions, seller info)

Pros: More data available Cons: Account gets suspended within days/weeks

3. GraphQL API mirroring

• Intercept GraphQL requests from real browser session
• Replay with matching headers (Authorization, X-FB-*, query hashes)
• Use residential proxies

Pros: Fastest, most data Cons: Extremely brittle, headers/hashes change constantly, high ban risk

Estimated Volume

• Montreal rental listings: 5,000-15,000+ active at any time
• New listings/day: High (possibly 100-500/day)
• Overlap with Kijiji: Moderate — some cross-posting, but many FB-only landlords

Legal Risk

High.

• Facebook ToS explicitly prohibits automated data collection
• Meta actively detects and blocks scrapers
• Meta has sued scraping companies (e.g., BrandTotal, Octopus)
• Account suspension is near-certain for logged-in scraping
• IP blocking happens for aggressive patterns

Comparison to Kijiji

Factor	Kijiji	Facebook Marketplace
Anti-bot	Moderate	Extreme
Login needed	No	No (basic), Yes (contact info)
Phone numbers	Available via API	Not available without login
DOM stability	Stable (__NEXT_DATA__)	Unstable (frequent changes)
Ban risk	Low	High
Data format	Structured JSON	Unstructured, obfuscated DOM
Maintenance effort	Low	Very high (constant fixes)
Implementation time	2-3 weeks	1-2 weeks + ongoing maintenance
Data richness	Very high	Low (without login)

Recommendation

Not recommended as a near-term priority.

Reasons:

1. No phone numbers without login → listings can't be actionable on our platform
2. Extreme anti-bot = high maintenance cost
3. Fragile selectors = scraper breaks frequently
4. Account ban risk if using login
5. Legal risk from Meta is real (they've sued scrapers)

When it might make sense:

• After Kangalou and other easy sources are done
• If a reliable third-party scraping service (Apify, Bright Data) handles the anti-bot
• If Facebook ever opens an API (unlikely)
• As a supplementary source for volume, accepting limited data quality

Better alternatives for the same effort:

• Kangalou (3-5 days, HTTP-only, 7,500 listings)
• Craigslist Montreal (simple, low volume)
• Louer.ca (Quebec-focused)