How HireFeed indexes, verifies, and ranks.

// 01The five principles.

Every decision downstream — what we index, how we rank, what we refuse to sell — traces back to one of these:

• // Principle 1
  Candidates never pay. Not for browsing, not for applying, not for the upcoming HireFeed Vetted credential. Revenue comes from buyers, never candidates.
• // Principle 2
  Pay is disclosed or the listing is out. If a job posting doesn't carry a rate, a range, or a salary band, we don't index it. This is the freshness moat — and the trust one.
• // Principle 3
  Candidate data is never sold. Not to platforms. Not to advertisers. Not to recruiters. Full stop, no exceptions, no "with your consent" workarounds.
• // Principle 4
  Rankings are methodology-published. Whether we're ranking the feed, ranking platforms, or reporting pay bands — the calculation is on this page. Sponsored placement, if it ever exists, is labeled inline.
• // Principle 5
  Corrections are public. If we get a number wrong, we change it and dated a note. If a platform contests a claim in "is-it-legit," we publish the response.

// 02How we index.

HireFeed's feed is aggregated continuously from a curated set of source platforms and lab careers pages. Three properties matter:

Sources

We index public job listings from a curated set of AI training platforms (Outlier, Mercor, Surge, Micro1, Alignerr, DataAnnotation, Handshake AI, and others), the careers pages of frontier AI labs, and reputable general-purpose freelance boards for Tier 2. Sources are added only when they meet pay-disclosure standards. We do not index scraped-from-scraper aggregators.

Freshness

Each source is re-polled on a per-source cadence — the fastest cycle is 60 seconds for high-velocity boards, up to 6 hours for slow-moving corporate careers pages. Closed roles drop within minutes of their close-time appearing on the source. The freshness stamp on each listing is real.

What we don't index

Listings without disclosed pay. Listings that fail our fraud classifier (payment-in-crypto requirements, WhatsApp-only communication, upfront-fee patterns). Listings whose source is a scraped aggregator rather than the primary source.

// 03Pay disclosure — what counts.

The word "pay-disclosed" is the promise. Here's what it means precisely, in decreasing order of specificity:

• // Tier A · Best
  An exact hourly rate or a tight rate band ($60–$75/hr). Approximately 40% of Tier-1 listings.
• // Tier B · Acceptable
  A rate range whose upper and lower bound differ by less than 3x ($40–$120/hr). Roughly 45% of Tier-1 listings.
• // Tier C · Salary band
  For full-time corporate roles (Tier 3): a salary band with equity split noted. Approximately 100% of Tier-3 US listings post-SB-1162 compliance.
• // Not counted
  "Competitive comp," "market rate," "DOE," "based on experience," or a range larger than 3x (e.g. $20–$200/hr with no domain qualifier). These postings are indexed as Pay: undisclosed and excluded from the Pay Index.

// 04Deduplication.

The same underlying role is often listed on multiple platforms — a lab posts to Mercor, Outlier, and Surge, sometimes to its own careers page as well. We deduplicate on three signals: hiring entity + task-skill signature + posting fingerprint. When a role is reposted, we retain the earliest version and note the alternative platforms in a "also listed on" field. This means our listing counts are lower than a naïve sum across sources, and that's the correct number.

// 05Ranking — how the feed is sorted.

The feed's default sort is a single scoring function:

• // tier_weight
  1.0 for AI training (Tier 1), 0.6 for freelance (Tier 2), 0.35 for US corporate (Tier 3). The track toggle on the homepage overrides this by pinning a single tier.
• // freshness_decay
  Exponential decay with half-life 24h for Tier 1, 72h for Tier 2, 7 days for Tier 3. Tier-1 markets move in hours.
• // pay_percentile
  Percentile of the listing's midpoint rate within its domain × task type. Higher-paid listings surface higher within their domain.
• // source_quality
  A stability score per source based on payment reliability signals and dispute history. Not a subjective platform ranking — a mechanical stability score.
• // fraud_penalty
  Rare, but if a listing carries any red-flag signal (payment-in-crypto, upfront-fee, WhatsApp-only), it's excluded, not down-ranked.

No sponsored ranking exists. If we ever accept sponsored placement, sponsored listings will carry an inline "Sponsored" label and be excluded from the ranking formula above.

// 06The Pay Index methodology.

The HireFeed Pay Index is built from Tier-A and Tier-B pay-disclosed listings only. Method:

• // Sample
  Trailing 30 days of pay-disclosed listings, deduplicated as in §04.
• // Typical range
  Middle 50% (25th–75th percentile) of midpoint hourly rates in the domain × task-type bucket.
• // Top decile
  90th percentile midpoint hourly rate.
• // 90-day trend
  Percent change of the median midpoint vs. the prior 90-day window.

During launch preview, sample sizes per row are not yet published. The production Pay Index will show sample-count and update-date per row.

// 07Platform comparisons.

The /alternatives/*, /is-it-legit/*, /pay/*, and /how-to-pass/* franchises follow the same rules:

Comparison tables

Ordering is derived from a scored rubric (pay disclosure, breadth of task types, queue reliability signals, worker sentiment on public forums). Not editorial preference. When rankings change, the change and its cause are noted at the top of the page.

Trust checklists (/is-it-legit)

Six evidence criteria per platform: payment history, corporate identity, task legitimacy, no candidate fees, queue reliability, screening feedback. Each is marked ✓ (verified), △ (mixed), or ✗ (failing) based on public evidence. If a platform disputes a rating, their response is published inline with the rating.

Sponsorship policy

We do not currently accept sponsorship from any platform we compare or rank. If that ever changes, the affected rating cannot be a ✓ upgrade tied to sponsorship. Any sponsored placement will be labeled inline and excluded from the ranking formula.

// 08What we won't do.

Some choices are more informative than the ones we make. HireFeed will not:

• // Won't 01
  Charge candidates any fee for any part of the product, ever.
• // Won't 02
  Sell, rent, license, or share candidate data outside the applicant's own explicit application to a specific listing.
• // Won't 03
  Accept sponsorship in exchange for higher ranking, better rating, or removal of an "is-it-legit" negative signal.
• // Won't 04
  Publish fabricated case-study numbers or invented retention metrics. Every claim on the site is defensible or clearly labeled illustrative.
• // Won't 05
  Claim a named competitor as a sourcing partner, credential-accepting party, or endorser without explicit written confirmation.

// 09Corrections & disputes.

If you find a mistake on this page, in the Pay Index, in a platform comparison, or anywhere else on the site — tell us. Corrections are shipped within 5 business days for factual errors and 15 business days for methodology changes.

Contact: corrections@hirefeed.co.in — include the URL, the specific claim, and the correction with a source if available. If you're a platform disputing an "is-it-legit" rating, include your evidence and we'll publish your response inline with the rating.