Google Hotels Scraper & Price Intelligence

Turn a destination into a clean hotel Dataset with public nightly prices, ratings, review counts, property identifiers, and explainable price intelligence. Enter only a destination; no Google API key, login, private token, paid API, or residential proxy is required.

What you get

• Valid, deduplicated hotel records from public Google Travel data.
• Nightly price and the currency actually displayed by Google.
• Rating, review count, star class, coordinates, property ID, and Google Travel URL when available.
• Median and price range when at least three comparable prices exist.
• Deterministic price_position, value_score, deal_signal, and coverage_confidence.
• Typed Dataset, machine-readable OUTPUT, and a responsive REPORT.html.
• Clean empty or partial results when Google returns no usable evidence.

Typical uses include hotel rate research, destination price snapshots, travel market analysis, accommodation lead lists, and comparison of public hotel positioning.

Quick start

The smallest useful input is:

{
  "destination": "Miami"
}

If dates are omitted, every run calculates them dynamically:

• checkIn: run date + 14 days
• checkOut: check-in + 1 day

The ADWIC/default input uses Miami, 2 adults, and 10 results without hard-coded dates:

{
  "destination": "Miami",
  "adults": 2,
  "maxResults": 10
}

An explicit example:

{
  "destination": "Madrid",
  "adults": 2,
  "currency": "EUR",
  "maxResults": 20
}

The prefilled and sample inputs never embed a fixed date. Add future ISO dates only when a specific stay window is required.

Input

Field	Required	Default	Limits
destination	Yes	Miami in the UI	2–120 characters
checkIn	No	Today +14 days	ISO YYYY-MM-DD, not in the past
checkOut	No	Check-in +1 day	After check-in; maximum 30 nights
adults	No	2	1–10
currency	No	Google-displayed currency	Three-letter ISO request
maxResults	No	20	1–50

Google can localize currency by network location even when a currency is requested. The Actor never relabels a price: it returns the observed currency and emits requested_currency_not_honored when appropriate.

Dataset fields

Stable public fields are returned when present: hotel_name, price_per_night, currency, rating, review_count, star_rating, latitude, longitude, property_id, and google_travel_url.

Detail-only or volatile fields are opportunistic: address, phone, provider_prices, availability, room, cancellation, and promotion. Missing evidence is represented as null, never guessed. Every row also includes destination, dates, source strategy, price intelligence, outlier state, and warnings.

Price intelligence

Price intelligence is computed only when at least three positive, same-result prices remain after outlier filtering.

1. Compute the preliminary median of positive nightly prices.
2. With at least three prices, mark values below 25% or above 400% of that median as outliers.
3. Exclude those outliers from the final benchmark.
4. Require at least three remaining prices; otherwise median, Value Score, price position, and Deal Signal are N/A/null.

For a comparable hotel:

• price ratio = hotel price / median price
• price component = clamp(50 + (1 − price ratio) × 100, 0, 65)
• quality component = rating / 5 × 25, or 0 when rating is unavailable
• review component = min(log10(review count + 1) / 4, 1) × 10, or 0 without reviews
• value_score = price component + quality component + review component, capped at 100

price_position is Below median at ≤90%, Near median between 90% and 110%, and Above median at ≥110%.

deal_signal rules:

• Strong: price ≤85% of median, rating ≥4.0, and at least 20 reviews
• Expensive: price >115% of median
• Fair: sufficient evidence but neither Strong nor Expensive
• N/A: insufficient sample, missing price, or price outlier

No “best value” claim is produced from an insufficient sample.

Acquisition and resilience

The Actor uses one public architecture:

Input → HTTP acquisition → selective browser fallback → normalization → identity/dedup → price intelligence → Dataset + OUTPUT + REPORT

• Direct HTTP structured acquisition is attempted first with a 25-second timeout per isolated alternative.
• Three safe Google Travel URL variants are attempted independently.
• CAPTCHA/unusual-traffic pages and HTTP 403/429 responses are rejected.
• The browser fallback has a 35-second cap and is used only when HTTP is empty or cannot verify requested date/currency state.
• If browser acquisition fails, usable HTTP evidence is retained with explicit warnings.
• If every strategy fails, the run returns a clean empty result without fabricated hotels.
• Worst-case acquisition timeout is bounded to approximately 110 seconds before normal output work.

Hotels are deduplicated first by Google property ID and then by normalized name plus coordinates/address. Explicit vacation rentals, holiday homes, and entire-apartment records are excluded when source evidence identifies them.

OUTPUT and REPORT

OUTPUT summarizes status, query, destination, dates, hotels found, valid price sample, median/min/max, observed currency, coverage confidence, warnings, report URL, acquisition attempts, and identity statistics.

REPORT.html provides a responsive destination summary, price range, hotel table, ratings, Value Scores, Deal Signals, Google Travel links, methodology, and warnings.

Billing preparation

The prepared pay-per-event name is hotel-result. A billing event is emitted only after a valid deduplicated hotel has been written to the Dataset and outputs have been delivered. Errors, timeouts, discarded candidates, and empty results are never billable.

The commercial event price is intentionally unset in this candidate. It must be chosen after the first Cloud cost/speed smoke and before commercial publication.

Real acquisition evidence

Research performed on 2026-09-10 recovered public data without a proxy for Miami, New York, and Madrid. Each raw HTTP capture contained 18 organic property records; browser verification displayed 20 accommodation cards per destination and future-date prices. The deterministic source captures, normalized fixtures, hashes, and browser observations are included under tests/fixtures.

Limitations

• Google Travel structured payloads are undocumented and may change.
• Ranking, availability, and provider prices are volatile.
• Requested currency may be replaced by IP-localized currency.
• Exact address and booking-provider offers may require additional browser/detail evidence and can remain null.
• A small no-proxy sample cannot guarantee that scaled traffic will never be rate-limited.
• This Actor is not affiliated with or endorsed by Google. Users are responsible for compliant usage.